News

The following paper has been accepted at Fragile Earth Workshop at KDD2020:

"One Forest: Towards a Global Species Dataset by Fusing Remote Sensing and Citizen Science Data with Graph Neural Networks" by Kenza Amara, David Dao, Björn Lütjens (MIT), Dava Newman (MIT), Tom Crowther (ETH), Ce Zhang.

A blog post entiteld "RumbleML, a declarative machine learning framework" by Can Berker Cikis has been posted on our Systems Group Blog.

The following Demo has been accepted at the 46th International Conference on Very Large Data Bases (VLDB 2020): 

"Ease.ml/snoopy in Action: Towards Automatic Feasibility Analysis for Machine Learning Application Development" by Cedric Renggli (ETH Zurich), Luka Rimanic (ETH Zurich), Luka Kolar ( ETH Zurich), Wentao Wu (Microsoft Research), Ce Zhang (ETH).

Abstract:

We demonstrate ease.ml/snoopy, a data analytics system that performs feasibility analysis for machine learning (ML) applications before they are developed. Given a performance target of an ML application (e.g., accuracy above 0.95), ease.ml/snoopy provides a decisive answer to ML developers regarding whether the target is achievable or not. We formulate the feasibility analysis problem as an instance of Bayes error estimation. That is, for a data (distribution) on which the ML application should be performed, ease.ml/snoopy provides an estimate of the Bayes error -- the minimum error rate that can be achieved by any classifier. It is well-known that estimating the Bayes error is a notoriously hard task. In ease.ml/snoopy we explore and employ estimators based on the combination of (1) nearest neighbor (NN) classifiers and (2) pre-trained feature transformations. To the best of our knowledge, this is the first work on Bayes error estimation that combines (1) and (2). In today's cost-driven business world, feasibility of an ML project is an ideal piece of information for ML application developers -- ease.ml/snoopy plays the role of a reliable ''consultant''.

The following papers have been accepted for presentation at the International Conference on Field-Programmable Logic and Applications (FPL) which will take place in August/September 2020:

"HyperLogLog Sketch Acceleration on FPGA" by Amit Kulkarni, Monica Chiosa, Thomas Preusser, Kaan Kara, David Sidler, Gustavo Alonso

"High Bandwidth Memory on FPGAs: A Data Analytics Perspective" by Kaan Kara, Gustavo Alonso, Christoph Hagleitner, Dionysios Diamantopoulos, Dimitris Syrivelis

"Using DSP Slices as Content-Addressable Update Queues" by Thomas Preusser, Monica Chiosa, Alexander Weiss, Gustavo Alonso

 The following paper has been accepted at SIGKDD 2020: 

"Building Continuous Integration Services for Machine Learning" by: Bojan Karlaš, Matteo Interlandi, Cedric Renggli, Wentao Wu, Ce Zhang, Deepak Mukunthu Iyappan Babu, Jordan Edwards, Chris Lauren, Andy Xu and Markus Weimer.

Gustavo Alonso will lead a Xilinx Adaptive Compute Cluster (XACC) at ETH Zurich. The cluster will support novel research in high performance computing. More in D-INFK Spotlight...

The following paper has been accepted at EuroSys 2020:

"StRoM: Smart Remote Memory" by David Sidler (Microsoft, USA & ETH Zurich), Zeke Wang (Zhejiang University, China & ETH Zurich), Monica Chiosa, Amit Kulkarni, Gustavo Alonso (ETH Zurich)

The paper "Lambada: Interactive Data Analytics on Cold Data using Serverless Cloud Infrastructure" by Ingo Müller (ETHZ), Renato Marroquín (ETHZ), Gustavo Alonso (ETHZ). has been accepted at SIGMOD 2020.

Gustavo Alonso gave a talk on "How the cloud is driving software and hardware specialization - a use case from the airline industry" at the EE Distinguished Speakers Seminar Series of IEL-EPFL.

Karolis Kusas joined the Systems Group as our new PhD Student.  Karolis completed his MSc at The University of Oxford.

The paper "Understanding video streaming algorithms in the wild" by Melissa Licciardello, Maximilian Grüner and Ankit Singla has been accepted at The Passive and Active Measurement (PAM) Conference in Oregon, USA, March 30-31, 2020.

The paper "Making Search Engines Faster by Lowering the Cost of Querying Business Rules Through FPGAs" authored by Fabio Maschi, Muhsen Owaida, Gustavo Alonso, Matteo Casalino (Amadeus), and Anthony Hock-Koon (Amadeus) has been accepted at SIGMOD 2020.

The following paper has been accepted at the 12th International Conference on Language Resources and Evaluation (LREC):  

"ZuCo 2.0: A Dataset of Physiological Recordings During Natural Reading and Annotation” by Nora Hollenstein (ETH Zürich), Marius Troendle (University of Zurich), Ce Zhang (ETH Zürich), Nicolas Langer (University of Zurich).

Shuai Zhang joined the Systems Group as our new PostDoc. Shuai comes to us from the University of New South Wales, Sydney, Australia.

Kaan Kara defended his PhD Dissertation "Specialized Hardware Solutions for In-Database Analytics and Machine Learning".

David Dao gave the following talk at the Applied Mashine Learning Days at EPFL. 

Scaling Natural Climate Solutions with Machine Learning

Abstract:

Nature-based Solutions (NbS) such as forests have the potential to deliver up to a third of the emission reduction required to achieve the Paris Agreement and limit climate change to safe levels. Yet, deforestation rates in Brazil are reaching its highest in decades and it is estimated that humanity has already cut down half of the world's forests. To date NbS receive less than 3% of available climate funding. In this talk we introduce Komorebi, a joint research project between ETH Zurich and the Government of Chile, that aims to protect, restore and fund NbS using machine learning systems. We leverage unsupervised learning and data programming on satellite and drone imagery to improve deforestation warning alerts and increase the efficiency of results-based payments for ecosystem services.

Vasia Kalavri joined Boston University as an Assistant Professor. John Liagouris joined Boston University as an Adjunct Assistant Professor. John also joined the Hariri Institute for Computing as a research scientist. 

Ingo Müller gave a talk entitled "Lambada: Interactive Data Analytics on Cold Data using Serverless Cloud Infrastructure" at Snowflake in Berlin. Technical Report

The paper “Network Topology Design at 27,000 Km/hour” by Debopam Bhattacherjee and Ankit Singla (ACM CoNEXT 2019) has been awarded with the Applied Networking Research Prize by IETF/IRTF for 2020.

Ioana Giurgiu, Oriana Riva, Dejan Juric, Ivan Krivulev, and Gustavo Alonso have been awarded The Test-of-Time Award of the ACM/IFIP/Usenix International Middleware Conference 2019 for their paper “Calling the Cloud: Enabling mobile phones as interfaces to cloud applications", published in Middleware 2009. Ioana Giurgiu, now a researcher at IBM Rüschlikon, gave a talk on the topic at Middleware 2019 (slides).

The following paper has been accepted at The First International Workshop on Deep Learning on Graphs: Methodologies and Applications (DLGMA’20) which will be held in conjuction with The Thirty Forth AAAI Conference on Artificial Intelligence 2020 in New York, NY, USA, February 7-12, 2020:

"An Anatomy of Graph Neural Networks Going Deep via the Lens of Mutual Information: Exponential Decay vs. Full Preservation" by Nezihe Merve Gürel (ETH Zurich), Hansheng Ren (Microsoft Research), Yujing Wang (Microsoft Research), Hui Xue (Microsoft Research), Yaming Yang (Microsoft Research) and Ce Zhang (ETH Zurich).

Ghislain Fourny gave a talk on Non-Nashian Game Theory at the Chair of Learning Sciences and Higher Education (D-GESS).

Fabio Maschi gave a talk entitled "An NFA-Based Approach for Accelerating Business Rule Inference on Reconfigurable Hardware" at Amadeus in Sophia Antipolis, France.

Timothy Roscoe gave an invited talk entitled "A fork() in the road" at Bell Labs at the celebration of the 50th Anniversary of Unix.

Read more in D-INFK spotlight story...

The following two papers have been accepted at Thirty Fourth AAAI Conference on Artificial Intelligence, February 7-12, 2020, New York, NY, USA.

"TextNAS: A Neural Architecture Search Space tailored for Text Representation" by Yujing Wang (MSRA), Yaming Yang (MSRA), Yiren Chen (Peking University),Jing Bai (Microsoft), Ce Zhang (ETH), Guinan Su (University of Science and Technology of China), Xiaoyu Kou (Peking University), Yunhai Tong (Peking University), Mao Yang (MSRA), Lidong Zhou (MSRA).

"Efficient Automatic CASH via Rising Bandits" by Yang Li (Peking University/ Systems Group visiting student), Jiawei Jiang (ETH Zurich), Jinyang Gao (Alibaba Group), Yingxia Shao (BUPT), Ce Zhang (ETH), Bin Cui (Peking University).

Gustavo Alonso gave a talk on the ongoing research on Enzian and data processing on modern hardware at the 2019 High Performance Transaction Processing Workshop (HPTS) in Asilomar, California.

The following paper co-authored by Shaoduo Gan has been accepted at the Workshop on Systems for ML at NeurIPS 2019: 

"Distributed Asynchronous Domain Adaptation: Towards Making Domain Adaptation More Practical in Real-World Systems" by Shaoduo Gan, Akhil Mathur, Anton Isopoussu, Nadia Berthouze, Nicholas Lane, Fahim Kawsar.

The paper "Causal limit on quantum communication" by Robert Pisarczyk, Zhikuan Zhao, Yingkai Ouyang, Vlatko Vedral, and Joseph F. Fitzsimons has been published in Physical Review Letters.

Popular summary by Zhikuan Zhao: 

We exploit the causal structure of quantum mechanics to study the rates of communication over noisy quantum channels. As a technical framework, we introduce a measure for quantum causality that quantifies temporal correlations between ends of a communication channel. We then prove it to be a general upper bound on the capacity of the channel to transmit quantum information.

Our result illuminates an example of harnessing quantum causality in practice. Just as the discovery of quantum entanglement-based protocols over the decades have furnished our comprehension of locality and realism, this revelation in the operational meaning of causal structure on information processing tasks will inspire insight into the nature of time in quantum mechanics.

From an applicational point of view, the newly discovered general bound for quantum channel capacity has the advantage of being efficiently computable comparing to many previously known approaches. As the need for efficient estimates of information transmission over more prevalent and complex networks of quantum channels is ever-growing, we believe our result will provide invaluable additional toolkits to study realistic noisy quantum channels and hence contribute to the design and future realization of a large-scale quantum internet.

Luka Rimanic joined the Systems Group as our new PostDoc. Luka completed his PhD at University of Bristol, UK.

The following paper has been accepted at CoNEXT 2019 (15th International Conference on emerging Networking EXperiments and Technologies), Orlando, Florida, U.S. December 9-12, 2019.

"Network design at 27,000 km/hour" by Debopam Bhattacherjee (ETH Zürich) and Ankit Singla (ETH Zürich).

The following papers and two demos have been presented at VLDB 2019. In addition, a BIRTE  workshop (co-located with VLDB) paper, has been presented in Los Angeles, California.

VLDB 2019 Papers:

Accelerating Generalized Linear Models with MLWeaving: A One-Size-Fits-All System for Any-precision Learning Zeke Wang (ETH Zurich), Kaan Kara (ETH Zurich), Hantian Zhang (ETH Zurich), Gustavo Alonso (ETH Zurich), Onur Mutlu (ETH Zurich), and Ce Zhang (ETH Zurich)

ColumnML: Column-Store Machine Learning with On-The-Fly Data Transformation Kaan Kara (ETH Zurich), Ken Eguro (Microsoft), Ce Zhang (ETH Zurich), and Gustavo Alonso (ETH Zurich)

Megaphone: Latency-conscious state migration for distributed streaming dataflows Moritz Hoffmann (ETH Zurich), Andrea Lattuada (ETH Zurich), Frank McSherry (ETH Zurich), Vasiliki Kalavri (ETH Zurich), John Liagouris (ETH Zurich), and Timothy Roscoe (ETH Zurich)

Efficient Task-Specific Data Valuation for Nearest Neighbor Algorithms Ruoxi Jia (UC Berkeley), David Dao (ETH Zurich), Boxin Wang (Zhejiang University), Frances Ann Hubis (ETH Zurich), Nezihe Merve Gürel (ETH Zurich), Bo Li (University of Illinois at Urbana–Champaign), Ce Zhang (ETH Zurich), Costas J. Spanos (UC Berkeley), and Dawn Song (UC Berkeley)

VLDB 2019 DEMOS:

doppioDB 2.0: Hardware Techniques for Improved Integration of Machine Learning into Databases by Kaan Kara (ETH Zurich), Zeke Wang (ETH Zurich), Ce Zhang (ETH Zurich), and Gustavo Alonso (ETH Zurich)

Ease.ml/ci and Ease.ml/meter in Action: Towards Data Management for Statistical Generalization Cedric Renggli (ETH Zurich), Frances Ann Hubis (ETH Zurich), Bojan Karlaš (ETH Zurich), Kevin Schawinski (Modulos AG), Wentao Wu (Microsoft Research, Redmond), and Ce Zhang (ETH Zurich)

Workshop paper (BIRTE): 

FASTER State Management for Timely Dataflow Matthew J Brookes, Vasiliki Kalavri, John Liagouris (ETH Zurich)

Ghislain Fourny wrote a blog post entiteld "Building an inverted index on a large text collection with JSONiq" on the Systems Group Blog.

The paper "Strong consistency is not hard to get: Two-Phase Locking and Two-Phase Commit on Thousands of Cores" by Claude Barthels, Ingo Müller, Konstantin Taranov, Torsten Hoefler, and Gustavo Alonso exploring the implementation of 2PL and 2PC at large scales using MPI has been accepted to VLDB 2020.

The following paper co-authored by Cédric Renggli has been accepted at the International Conference for High Performance Computing, Networking, Storage and Analysis (SC’19) this year in Denver.

" SparCML: High-Performance Sparse Communication for Machine Learning " by Cedric Renggli, Saleh Ashkboos, Mehdi Aghagolzadeh, Dan Alistarh and Torsten Hoefler

Ce Zhang and Gustavo Alonso have presented work done at the systems group on Machine learning and hardware infrastructures for machine learning at the 20th Microsoft Research Faculty Summit in Redmond, Wa., USA.

Vasiliki Kalavri and her research highlights have been presented in the Department of Computer Science Spotlight article "Building a stream processor for the future".

The following paper co-authored by David Dao has been accepted at Poly'19 Workshop at VLDB 2019.

"Data Capsule: A New Paradigm for Automatic Compliance with Data Privacy Regulations" by Lun Wang (UC Berkeley), Joseph P. Near (University of Vermont), Neel Somani (UC Berkeley), Peng Gao (UC Berkeley), Andrew Low (UC Berkeley), David Dao (ETH Zurich), and Dawn Song (UC Berkeley).

Frank McSherry has been awarded the Test-Of-Time Award at SIGMOD 2019 for his paper "Privacy integrated queries: an extensible platform for privacy-preserving data analysis" published 10 years earlier at SIGMOD 2009.

The following paper co-authored by John Liagouris has been accepted for publication at the VLDB Journal:

"SRX: Efficient Management of Spatial RDF Data" by Konstantinos Theocharidis (University of Peloponnese/IMSI, RC 'Athena'), John Liagouris (ETH Zürich), Nikos Mamoulis (University of Ioannina), Panagiotis Bouros (Johannes Gutenberg University Mainz), Manolis Terrovitis (IMSI, RC 'Athena')

Abstract:

We present a general encoding scheme for the efficient management of spatial RDF data. The scheme approximates the geometries of the RDF entities inside their (integer) IDs and can be used, along with several operators and optimizations we introduce, to accelerate queries with spatial predicates and to re-encode entities dynamically in case of updates. We implement our ideas in SRX, a system built on top of the popular RDF-3X system. SRX extends RDF-3X with support for three types of spatial queries: range selections (e.g. find entities within a given polygon), spatial joins (e.g. find pairs of entities whose locations are close to each other), and spatial k nearest neighbors (e.g. find the three closest entities from a given location). We evaluate SRX on spatial queries and updates with real RDF data, and we also compare its performance with the latest versions of three popular RDF stores. The results show SRX’s superior performance over the competitors; compared to RDF-3X, SRX improves its performance for queries with spatial predicates while incurring little overhead during updates.

Gustavo Alonso gave a keynote presentation "How Hardware Evolution is Driving Software Systems" at the 13th ACM International Conference on Distributed and Event-Based Systems, DEBS, in Darmstadt, Germany.

The following paper has been accepted at ICML Climate Change Workshop at 36th International Conference on Machine Learning, Long Beach, California, 2019. 

GainForest: Scaling Climate Finance for Forest Conservation using Interpretable Machine Learning on Satellite Imagery by David Dao, Catherine Cang, Clement Fung, Ming Zhang, Reuven Gonzales, Nick Beglinger, Ce Zhang.

Kaan Kara obtained third place in System Design Contest (FPGA category) of the 2019, 56th Design Automation Conference (DAC 2019). More than 100 international teams participated in the two categories (GPU and FPGA).

The project "SnowBell: An FPGA Computing Platform for Machine Learning in Data Analytics" by Mohsen Ewaida has been awarded an ETH Pioneer Fellowship with financing from the ETH Foundation. The Fellowships are intended to help with the development of a highly innovative product or service to be exploited commercially.

The following paper has been published in Springer Quantum Machine Intelligence:

Bayesian deep learning on a quantum computer by Zhikuan Zhao, Alejandro Pozas-Kerstjens, Patrick Rebentrost, Peter Wittek.

Maximilian Grüner joined the Systems Group as our new PhD student. Maximilian completed hid MSc at ETH.

MIT Technology review published an article entitled "How AI could save lives without spilling medical secrets" on David Dao's work on Kara, a collaboration with Berkeley and Stanford Medical School.

Jiawei Jiang joined the Systems Group as our new PostDoc. Jiawei completed his PhD at Peking University.

Anastasiia Ruzhanskaia joined the Systems Group as our new PhD student. Anastasiia is coming to us from Moscow Institute of Physics and Technology.

Timothy Roscoe talks about a research computer, Enzian, and shares more about his research in the D-INFK spotlight article "Bridging the gap: a computer built for research".

Moritz Hoffmann defended his PhD Dissertation “Managing and understanding distributed stream processing”.

Moritz will join Google Zürich in June 2019.

The following paper has been accepted at VLDB 2019, Los Angeles, CA, August 26-30, 2019.

"Megaphone: Latency-conscious state migration for distributed streaming dataflows" by Moritz Hoffmann, Andrea Lattuada, Frank McSherry, Vasiliki Kalavri, John Liagouris, Timothy Roscoe.

Timothy Roscoe gave a keynote address entitled "Building Enzian, a research computer" at the 9th Workshop Systems for Future Multicore Architectures (SFMA) in Berlin.

Cedric Renggli presented the following paper at SysML'19, Stanford, CA, USA, March 31-April 2, 2019:

"Continuous Integration of Machine Learning Models: A Rigorous Yet Practical Treatment" by Cedric Renggli (ETH Zurich), Bojan Karlaš (ETH Zürich), Bolin Ding ("Data Analytics and Intelligence Lab, Alibaba Group"), Feng Liu (Huawei Technologies), Kevin Schawinski (Modulos AG), Wentao Wu (Microsoft Research), Ce Zhang (ETH Zurich).

The article "How Artificial Intelligence Is Changing Science" about Ce Zhang's astronomy work with Kevin Schawinski has been published in Quanta Magazine.

Zhenhao He joined the Systems Group as our new PhD student. Zhenhao completed his Master Thesis in the Systems Group. 

John Liagouris gave the following talk at eBay, San Jose, CA.

Strymon: Providing fast and meaningful insights into enterprise datacenters

Abstract:

In this talk I will present an overview of the Strymon project at ETH Zurich. The project focuses on the design and development of Strymon, a system that leverages existing datacenter logging pipelines to ingest and process logs of events in real time and provide datacenter administrators with timely insights into the functionality of the running systems. Strymon is written in Rust and builds on top of timely dataflow, a high-performance engine for parallel and distributed streaming computations. The talk will focus on two Strymon use cases: (i) online reconstruction of user sessions from individual logs, which is often the first step in many datacenter management tasks, and (ii) online critical path analysis of long-running applications, which can be used to identify performance bottlenecks at runtime.

The paper "Entity Recognition at First Sight: Improving NER with Eye Movement Informationby" by Nora Hollenstein and Ce Zhang has been accepted at the Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL 2019), June 2–7, 2019 in Minneapolis, USA.

Tom Anderson from the University of Washington joined the Systems Group as our Visiting Professor until June 2019.

The paper "Accelerating Generalized Linear Models with MLWeaving: A One-Size-Fits-All System for Any-precision Learning" by Zeke Wang, Kaan Kara, Hantian Zhang, Gustavo Alonso, Onur Mutlu, and Ce Zhang has been accepted to VLDB'19.

Vasiliki Kalavri gave a talk at the "Tech talks on Scientific Computing" event at Ghent University organized by the HPC Group.

Title: Towards self-managed, re-configurable streaming dataflow systems

Slides
Video

Timothy Roscoe gave an invited talk "Enzian: research hardware for systems support for AI" at the ACM/Microsoft India Academic Research Summit.

Fabio Maschi joined the Systems Group as our new PhD student. Fabio completed his Master Thesis at Paris-Saclay University.

Gustavo Alonso gave a talk at Oracle on the use of RDMA for in-network data processing through FPGA based NICs.

.

Claude Barthels defended his PhD Dissertation “Scalable Query and Transaction Processing over High-Performance Networks”.

Gustavo Alonso gave the talk "How Hardware Evolution is Driving Software Systems" at the Department of Computer Science of McGill University, Montreal, Canada, covering the latest results from the group on FPGA research and developments around the Enzian project.

Zsolt István is one of the ETH Outstanding Doctoral Theses 2018 Award Winners for his PhD Thesis: "Building Distributed Storage with Specialized Hardware".  (Department of Computer Science Award Winners).

Abstract:

In an effort to keep up with increasing data sizes, applications in the datacenter are scaled out to large number of machines. Even though this allows them to tackle complex problems, data movement bottlenecks of various types appear, limiting overall performance and scalability. Pushing computation closer to the data reduces bottlenecks, and in this work we explore this idea in the context of distributed key-value stores.

Efficient compute is important in the datacenter, especially at scale. Driven by the stagnation of single-threaded CPU performance, specialized hardware and hybrid architectures are emerging and could hold the answer for more efficient compute and data management. In this work we use Field Programmable Gate Arrays (FPGA) to break traditional trade-offs and limitations, and to explore design scenarios that were previously infeasible.

This dissertation focuses on distributed storage, a building block in scale out applications. It explores how such a service can benefit from specialized hardware nodes. We focus in particular on providing complex near-data computation with the goal of reducing the data movement bottleneck between application layers. Furthermore, this work addresses a shortcoming in the design of most distributed storage nodes, namely the mismatch between computational power and network/storage media bandwidth.

The mismatch is present because, if regular server machines are used, there is plenty of processing power to implement various filtering and processing operations, but the overall architecture is over-provisioned compared to the network. In contrast, if specialized hardware nodes are used (e.g. network-attached flash) the internal and external bandwidths are better matched, but these nodes will not be able to carry out complex processing near the data without slowing data access down. Our solution, Caribou, proposes a balanced design point: small-footprint hardware nodes that, even though offer high throughput and low latency, are also flexible to adapt to different workloads and processing types without being over-provisioned.

The work presented in this dissertation is not a one-off effort: it provides an extensible and i modular architecture for storage nodes that can be used as a platform for implementing near-data processing ideas for various application domains. The lessons are be applicable for different storage media or networking technologies as well.

The following papers have been presented at MICRO, IMC and OSDI  in October 2018:

 • Yaohua Wang, Arash Tavakkol, Lois Orosa, Saugata Ghose, Nika Mansouri Ghiasi, Minesh Patel, Jeremie S. Kim, Hasan Hassan, Mohammad Sadrosadati, and Onur Mutlu, "Reducing DRAM Latency via Charge-Level-Aware Look-Ahead Partial Restoration" Proceedings of the 51st International Symposium on Microarchitecture (MICRO), Fukuoka, Japan, October 2018. [Lightning Talk Video] 

• Justin Meza, Tianyin Xu, Kaushik Veeraraghavan, and Onur Mutlu, "A Large Scale Study of Data Center Network Reliability" Proceedings of the 18th ACM Internet Measurement Conference (IMC), Boston, MA, USA, October/November 2018. 

• Kevin Hsieh, Ganesh Ananthanarayanan, Peter Bodik, Shivaram Venkataraman, Paramvir Bahl, Matthai Philipose, Phillip B. Gibbons, and Onur Mutlu, "Focus: Querying Large Video Datasets with Low Latency and Low Cost" Proceedings of the 13th USENIX Symposium on Operating Systems Design and Implementation (OSDI), Carlsbad, CA, USA, October 2018.

Rodrigo Bruno joined the Systems Group as our new Postdoc. Rodrigo completed his PhD at Instituto Superior Técnico, Lissabon, Portugal.

The paper 'Sequence Classification with Human Attention' by Maria Barrett, Joachim Bingel, Nora Hollenstein, Marek Rei and Anders Søgaard has won the 'Special award for the best paper on research inspired by human language learning and processing' at CoNLL 2018 (The SIGNLL Conference on Computational Natural Language Learning).

The paper "DPI: The Data Processing Interface for Modern Networks" by Gustavo Alonso, Carsten Binnig, Ippokratis Pandis, Kenneth Salem, Jan Skrzypczak, Ryan Stutsman, Lasse Thostrup, Tianzheng Wang, Zeke Wang, and Tobias Ziegler has been accepted for publication at CIDR 2019 which will take place in Asilomar, California, USA, in January 2019.

Vasiliki Kalavri presented the following paper at OSDI 2018 in Karlsbad, CA, USA:

"Three steps is all you need: fast, accurate, automatic scaling decisions for distributed streaming dataflows" by Vasiliki Kalavri, John Liagouris, Moritz Hoffmann, and Desislava Dimitrova, ETH Zurich; Matthew Forshaw, Newcastle University; Timothy Roscoe, ETH Zurich.

The paper "FPGA-based TCP/IP Checksum Offloading Engine for 100 Gbps Networks" by Mario Ruiz, Sergio Lopez, Gustavo Sutter, and Gustavo Alonso has been accepted at ReConFig 2018, and will be presented at the conference in Cancun, Mexico, on December 2018.

Gustavo Alonso gave an invited talk at KTH, Stockholm, Sweden, on "Hardware Acceleration Close to the Network" on the 26th of September, 2018.

The following paper has been accepted at the 17th ACM Workshop on Hot Topics in Networks (HotNets 2018).

"Gearing up for the 21st century space race" by Debopam Bhattacherjee (ETH Zürich), Waqar Aqeel (Duke University), Ilker Nadi Bozkurt (Duke University),  Anthony Aguirre (University of California, Santa Cruz), Balakrishnan Chandrasekaran (Max-Planck-Institut für Informatik), P. Brighten Godfrey (University of Illinois at Urbana-Champaign), Gregory P. Laughlin (Yale University),  Bruce M. Maggs (Duke University),  Ankit Singla (ETH Zürich).

Merve Gürel presented the following paper at the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshop:

"Towards More Accurate Radio Telescope Images" by Nezihe Merve Gürel, Paul Hurley and Matthieu Simeoni.

Abstract:

Radio interferometry usually compensates for high levels of noise in sensor/antenna electronics by throwing data and energy at the problem: observe longer, then store and process it all. We propose instead a method to remove the noise explicitly before imaging. To this end, we developed an algorithm that first decomposes the instances of antenna correlation matrix, the so-called visibility matrix, into additive components using Singular Spectrum Analysis and then cluster these components using graph Laplacian matrix. We show through simulation the potential for radio astronomy, in particular, illustrating the benefit for LOFAR, the low frequency array in Netherlands. Least-squares images are estimated with far higher accuracy with low computation cost without the need for long observation time.

Gustavo Alonso gave the talk "How Hardware Evolution is Driving Software Systems" at VMware, Palo Alto California on Sept. 6th during a visit to VMware Research.

Frances Ann Hubis joined the Systems Group as our new PhD student. Frances completed her Master Thesis at ETH.

Gustavo Alonso gave an invited talk at the 2nd Workshop on Reconfigurable Computing for Machine Learning (RCML'2018), August 30th, 2018, Dublin, Ireland: "What reconfigurable computing can do for machine learning".

Abstract:

Data Processing is undergoing a multitude of interesting changes: from the platforms (cloud, appliances) to the workloads, data types, and operations. Machine Learning has become the dominant workload in data processing, giving raise to many challenges. These challenges are being tackled through innovation in hardware even to the point of having fully specialized designs for particular applications. In this talk I will review some of the most important changes happening in hardware, with an emphasis on FPGAs, and discuss how they can be used in machine learning as well as the opportunities they create.

Several papers have been presented at FPL'18 on work done in the group around data processing on FPGAs:

Zsolt Istvan (Systems Group alumnus) presented the paper “Providing Multi-tenant Services with FPGAs: Case Study on a Key-Value Store” (by Zsolt Istvan, Gustavo Alonso and Ankit Singla).

Muhsen Owaida presented the paper “Application Partitioning on FPGA Clusters: Inference over Decision Tree Ensembles” (by Muhsen Owaida, Gustavo Alonso).

Zhenhao He presented the paper "A Flexible K-Means Operator for Hybrid Databases" (by Zhenhao He, David Sidler, Zsolt Istvan and Gustavo Alonso).

The following paper has been accepted at the SIGNLL Conference on Computational Natural Language Learning (CoNLL 2018 ), in Brussels, Belgium, October 31 - November 1, 2018:

"Sequence classification with human attention" by Maria Barrett, Joachim Bingel*, Nora Hollenstein**, Marek Rei***, Anders Søgaard*

* University of Copenhagen
** ETH Zurich
*** University of Cambridge, United Kingdom

Onur Mutlu gave an invited talk entitled "RowHammer and Beyond" at Microsoft Research 2018 Faculty Summit (MSR FACSUMMIT), Redmond, WA, USA, August 2018.

Slides (pptx) (pdf)

Abishek Ramdas joined the System Group today as PhD student. He comes to us from Qualcomm Technologies, San Diego, CA, USA.

The following paper has been accepted for publication at the Symposium of Cloud Computing (SoCC'18), October 11–13, 2018, Carlsbad, CA, USA

Pay One, Get Hundreds for Free: Reducing Cloud Costs through Shared Query Execution by Renato Marroquín, Ingo Müller, Darko Makreshanski (OracleLabs), Gustavo Alonso

Yishai Oltchik joined the Systems Group as a new PhD student. Yishai comes to us from Hebrew University of Jerusalem.

Onur Mutlu gave a keynote talk "Processing Data Where It Makes Sense in Modern Computing Systems: Enabling In-Memory Computation" at 7th Mediterranean Conference on Embedded Computing (MECO), Budva, Montenegro, June 2018.

 Slides

Onur Mutlu gave the keynote talk "Rethinking Memory System Design: Robustness, Energy, Performance" at the 24th IEEE International Symposium on On-Line Testing and Robust System Design (IOLTS) in July 2018. Slides

Simon Gerber defended his PhD Dissertation “Authorization, Protection, and Allocation of Memory in a Large System”.

Moritz Hoffmann presented the following paper at the Algorithms and Systems for MapReduce and Beyond Workshop (BeyondMR):

“Latency-conscious dataflow reconfiguration” by Moritz Hoffmann, Frank Mcsherry and Andrea Lattuada.

Gustavo Alonso received the Distinguished Alumnus Award from the Department of Computer Science at the University of California, Santa Barbara, where he obtained his PhD.

The following paper has been accepted at VLDB 2018 (Experiments and Analyses track):

Streaming Graph Partitioning: An Experimental Study by Z. Abbas, V. Kalavri, P. Carbone, V. Vlasso. 

Gustavo Alonso has authored a short survey of papers covering recent developments on FPGAs in data centers at ACM Queue, March/April 2018.

The following three papers have been accepted at the 28th International Conference on Field Programmable Logic & Applications (FPL 2018) in Dublin, Ireland, August 27 - 31, 2018:

“Application Partitioning on FPGA Clusters: Inference over Decision Tree Ensembles” by Muhsen Owaida, Gustavo Alonso.

"A Flexible K-Means Operator for Hybrid Databases" by Zhenhao He, David Sidler, Zsolt Istvan and Gustavo Alonso.

“Providing Multi-tenant Services with FPGAs: Case Study on a Key-Value Store” by Zsolt Istvan, Gustavo Alonso and Ankit Singla.

The following paper has been accepted at VLDB 2018 in Rio De Janeiro, Brasil, August 27-31, 2018:

"MLBench: Benchmarking Machine Learning Services Against Human Experts" by Yu Liu, Hantian Zhang, Luyuan Zeng, Wentao Wu (Microsoft Research, Redmond), Ce Zhang.

The following two papers have been accepted as demos at VLDB 2018 in Rio De Janeiro, Brasil, August 27-31, 2018:

"Ease.ml in Action: Towards Multi-tenant Declarative Learning Services" by Bojan Karlas, Ji Liu, Wentao Wu, Ce Zhang.

 "A Demonstration of Sterling: A Privacy-Preserving Data Marketplace" by Nick Hynes, Raymond Cheng, Noah Johnson, David Dao, Dawn Song.

Gustavo Alonso gave the talk "The impact of modern hardware on system design" at the 8th Workshop on Systems for Multi-core and Heterogeneous Architectures (SFMA 2018) and the talk "20+ Years of data replication and consistency. Have we learned anything?" at the 5th Workshop on Principles and Practice of Consistency for Distributed Data (PaPoC 2018), both collocated with EuroSys 2018, in Porto, Portugal.

---

"The impact of modern hardware on system design"

Abstract:

Computing Systems are undergoing a multitude of interesting changes: from the platforms (cloud, appliances) to the workloads, data types, and operations (big data, machine learning). Many of these changes are driven or being tackled through innovation in hardware even to the point of having fully specialized designs for particular applications. In this talk I will review some of the most important changes happening in hardware and discuss how they affect system design as well as the opportunities they create. I will focus on data processing as an example but also discuss applications in other areas.

---

"20+ Years of data replication and consistency. Have we learned anything?"

Abstract:

Data consistency is a fundamental topic in computer science, cutting across many areas and applications. Surprisingly, a significant part of ongoing research revolves around ideas pursued decades ago. The fact that published papers systematically ignore all this previous work does not mean it does not exist. Maybe the context has changed (cloud, geo-replication, massive scale, etc.) but the principles behind consistency for replicated data remain the same and have been known for a long time. In this talk, I will argue that there is a corpus of basic principles governing data consistency and that the mapping of such principles to mechanisms and system implications is also well established after decades of research. By providing a historical perspective of how data replication, and the subsequent problem of consistency, has evolved over the years, I will enumerate these principles, discuss the performance aspects associated to them, and relate these ideas to existing systems and trends. The goal of the talk is to open up new perspectives in an area where there is still much to be done and that has become highly relevant in the era of large scale computing infrastructures.

Ingo Müller presented the paper "Reproducible Floating-Point Aggregation in RDBMSs" (Ingo Müller, Andrea Arteaga, Torsten Hoefler, Gustavo Alonso) at the 34th IEEE International Conference on Data Engineering (ICDE2018) in Paris, April 2018.

The following three papers have been accepted at the the 45th International Symposium on Computer Architecture (ISCA 2018):

• Arash Tavakkol, Mohammad Sadrosadati, Saugata Ghose, Jeremie Kim, Yixin Luo, Yaohua Wang, Nika Mansouri Ghiasi, Lois Orosa, Juan G. Luna and Onur Mutlu, "FLIN: Enabling Fairness and Enhancing Performance in Modern NVMe Solid State Drives" to appear in Proceedings of the 45th International Symposium on Computer Architecture (ISCA), Los Angeles, CA, USA, June 2018.

• Nandita Vijaykumar, Eiman Ebrahimi, Kevin Hsieh, Phillip B. Gibbons and Onur Mutlu, "The Locality Descriptor: A Holistic Abstraction to Exploit Data Locality in Graphics Processing Units" to appear in Proceedings of the 45th International Symposium on Computer Architecture (ISCA), Los Angeles, CA, USA, June 2018.

• Nandita Vijaykumar, Abhilasha Jain, Diptesh Majumdar, Kevin Hsieh, Gennady Pekhimenko, Eiman Ebrahimi, Nastaran Hajinazar, Phillip B. Gibbons and Onur Mutlu, "A Case for Richer Cross-layer Abstractions: Bridging the Semantic Gap to Enhance Memory Optimization" to appear in Proceedings of the 45th International Symposium on Computer Architecture

The folowing paper has been accepted at the 9th International Conference on Interactive Theorem Proving (ITP'18) in Oxford, United Kingdom, July 9-12 2018:

"Physical addressing on real hardware in Isabelle/HOL" by Reto Achermann, Lukas Humbel, David Cock and Timothy Roscoe

Cedric Renggli joined the Systems Group as a new PhD student. Cedric completed his Master Theis in the Systems Group.

The following paper coauthored by Ce Zhang and Hantian Zhang has been published in Monthly Notices of the Royal Astronomical Society:

"PSFGAN: a generative adversarial network system for separating quasar point sources and host galaxy light" by Dominic Stark, Barthelemy Launet, Kevin Schawinski, Ce Zhang, Michael Koss, M Dennis Turp, Lia F Sartori, Hantian Zhang, Yiru Chen, Anna K Weigel.

Abstract

The study of unobscured active galactic nuclei (AGN) and quasars depends on the reliable decomposition of the light from the AGN point source and the extended host galaxy light. The problem is typically approached using parametric fitting routines using separate models for the host galaxy and the point spread function (PSF). We present a new approach using a Generative Adversarial Network (GAN) trained on galaxy images. We test the method using Sloan Digital Sky Survey (SDSS) r-band images with artificial AGN point sources added which are then removed using the GAN and with parametric methods using GALFIT. When the AGN point source PS is more than twice as bright as the host galaxy, we find that our method, PSFGAN, can recover PS and host galaxy magnitudes with smaller systematic error and a lower average scatter (49%). PSFGAN is more tolerant to poor knowledge of the PSF than parametric methods. Our tests show that PSFGAN is robust against a broadening in the PSF width of ±50% if it is trained on multiple PSF’s. We demonstrate that while a matched training set does improve performance, we can still subtract point sources using a PSFGAN trained on non-astronomical images. While initial training is computationally expensive, evaluating PSFGAN on data is more than 40 times faster than GALFIT fitting two components. Finally, PSFGAN it is more robust and easy to use than parametric methods as it requires no input parameters.

Claude Barthels gave the talk "Distributed Join Algorithms on Thousands of Cores" at Imperial College London and Microsoft Research Cambridge.

Onur Multu gave a keynote talk entitled "Accelerating Genome Analysis: A Primer on an Ongoing Journey" at Workshop on Accelerator Architecture in Computational Biology and Bioinformatics (AACBB), Vienna, Austria, February 2018. Slides

Timothy Roscoe gave the talk "Enzian: a research computer" at IBM Research in Rüschlikon.

The following paper, has been accepted at the ACM International Conference on Measurement and Modeling of Computer Systems (SIGMETRICS), Irvine, CA, USA, June 2018:

Saba Ahmadian, Onur Mutlu, and Hossein Asadi: "ECI-Cache: A High-Endurance and Cost-Efficient I/O Caching Scheme for Virtualized Platforms".

The following five papers have been accepted at the 23rd ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) , March 24th – March 28th, Williamsburg, VA, USA:  

• Rachata Ausavarungnirun, Vance Miller, Joshua Landgraf, Saugata Ghose, Jayneel Gandhi, Adwait Jog, Christopher J. Rossbach, and Onur Mutlu, "MASK: Redesigning the GPU Memory Hierarchy to Support Multi-Application Concurrency"

• Maciej Besta, Syed Minhaj Hassan, Sudhakar Yalamanchili, Rachata Ausavarungnirun, Onur Mutlu, Torsten Hoefler, "Slim NoC: A Low-Diameter On-Chip Network Topology for High Energy Efficiency and Scalability".

• Amirali Boroumand, Saugata Ghose, Youngsok Kim, Rachata Ausavarungnirun, Eric Shiu, Rahul Thakur, Daehyun Kim, Aki Kuusela, Allan Knies, Parthasarathy Ranganathan, and Onur Mutlu, "Google Workloads for Consumer Devices: Mitigating Data Movement Bottlenecks".

• Amir M. Rahmani, Bryan Donyanavard, Tiago Mück, Kasra Moazzemi, Axel Jantsch, Onur Mutlu, and Nikil Dutt, "SPECTR: Formal Supervisory Control and Coordination for Many-core Systems Resource Management".

• Mohammad Sadrosadati, Amirhossein Mirhosseini, Seyed Borna Ehsani, Hamid Sarbazi-Azad, Mario Drumond, Babak Falsafi, Rachata Ausavarungnirun, and Onur Mutlu, "LTRF: Enabling High-Capacity Register Files for GPUs via Hardware/Software Cooperative Register Prefetching".

David Dao and his team won a Microsoft AI for Earth Award for their deforestation prediction project. The project is using satellite imagery to predict the spread of deforestation over the Amazon with deep learning and is supported by Microsoft and Esri.  http://gainforest.org

Nora Hollenstein, our PhD student, and Jonathan Rotsztejn, our Master Thesis student,  won the 2017 International Workshop on Semantic Evaluation (SemEval-2017) "relation classification" competition, placing first out of 28 teams from all over the world.

Definition on semantic evaluation

Our alumnus, Tim Kraska, joins MIT (Computer Science & Artificial Intelligence Lab - CSAIL) as an Associate Professor. MIT News

Zsolt István defended his PhD Dissertation "Building Distributed Storage with Specialized Hardware".

The paper "R-OSGi: Distributed Applications through Software Modularization", published at the ACM/IFIP/Usenix International Middleware Conference 2007, has received the Test-of-Time Award at the 2017 Middleware Conference. Jan Rellermeyer, now professor at TU Delft, received the award and gave a talk on the evolution of R-OSGi.

The following paper has been accepted at the 15th USENIX Symposium on Networked Systems Design and Implementation (NSDI '18) in Renton, WA, USA, April 9-11, 2018:

SnailTrail: Generalizing Critical Paths for Online Analysis of Distributed Dataflows by Moritz Hoffmann, Andrea Lattuada, John Liagouris, Vasiliki Kalavri, Desislava Dimitrova, Sebastian Wicki, Zaheer Chothia, Timothy Roscoe.

Onur Mutlu gave the Distinguished Lecture "Rethinking Memory System Design (and the Computing Platforms We Design Around It)" at INESC-ID in Lisbon, Portugal. Slides

Abstract:

The memory system is a fundamental performance and energy bottleneck in almost all computing systems. Recent system design, application, and technology trends that require more capacity, bandwidth, efficiency, and predictability out of the memory system make it an even more important system bottleneck. At the same time, DRAM and flash technologies are experiencing difficult technology scaling challenges that make the maintenance and enhancement of their capacity, energy efficiency, and reliability significantly more costly with conventional techniques. In fact, recent reliability issues with DRAM, such as the RowHammer problem, are already threatening system security and predictability.

In this talk, we first discuss major challenges facing modern memory systems in the presence of greatly increasing demand for data and its fast analysis. We then examine some promising research and design directions to overcome these challenges. We discuss three key solution directions: 1) enabling new memory architectures, functions, interfaces, via more memory-centric system design, 2) enabling emerging non-volatile memory (NVM) technologies via hybrid and persistent memory systems, 3) enabling predictable memory systems via QoS-aware memory system design. If time permits, we will also discuss research challenges and opportunities in NAND flash memories.

The XCell Daily Blog of Xilinx has highlighted the work done by Kaan Kara implementing the Zip-ML framework of Prof. Ce Zhang into an FPGA. More...

Timothy Roscoe gave a Colloquium entitled “The Trouble with Hardware” at Pierre-and-Marie-Curie University (UPMC) in Paris, France.

Abstract:

Computer hardware, from datacenters through rackscale computing down to mobile device systems-on-chip, is increasingly easy to design. A combination of advanced CAD systems, the rise of Moore's law, and now the fall of Moore's law, has resulted in a huge diversity of hardware platforms whose complexity is immense. One downside of this is the monumental software engineering challenge in building and maintaining correct, robust, and portable systems software. This is an open secret in many pockets of industry but receives little attention in research. We ran against this problem full square (and continue to do so) while developing the Barrelfish research OS. In part of my talk I'll discuss about what can be done to address this in the design of systems software, by importing ideas from formal verification, knowledge representation, and program synthesis to the C-dominated world of low-level code. This, however, begs broader questions: given that custom hardware is getting easier to design, what should it look like to system software? How can systems researchers influence such hardware? And in an age where more corporations are building custom hardware but academia is mostly restricted to commodity systems, what can be done to conduct relevant, impactful research in this space outside of industry? I'll try and suggest some answers.

The following two papers have been accepted at the 21st International Conference on Extending Database Technology (EDBT 2018), March 26-29, 2018 in Vienna, Austria.

MTBase : Optimizing Cross-Tenant Database Queries by Lucas Braun, Renato Marroquín, Kai-En Tsay, Donald Kossmann.

Synchronous Multi-GPU Deep Learning with Low-Precision Communication: An Experimental Study by Demjan Grubic (our master's student), Leo Tam (NVIDIA), Dan Alistarh, Ce Zhang.

Jana Giceva is one of the ETH Outstanding Doctoral Theses 2017 Award Winners for her PhD Thesis: Database/Operating system co-design.

Timothy Roscoe gave the talk "Intelligently Diagnosing Datacenters" at Huawei's Corporate Reliability Conference in Shenzhen, China.

Onur Mutlu gave the keynote talk entitled “"Processing Data Where It Makes Sense: Enabling In-Memory Computation" at the Mobile System Technologies (MST) 2017 Workshop.

Slides: PDF

Melissa Licciardello joined the Systems Group as a PhD student. Melissa completed her Master Thesis at Bologna University, Italy. 

The following four papers have been presented at the 50th Annual IEEE/ACM International Symposium on Microarchitecture (MICRO 50) in Boston, MA, USA.

• Vivek Seshadri, Donghyuk Lee, Thomas Mullins, Hasan Hassan, Amirali Boroumand, Jeremie Kim, Michael A. Kozuch, Onur Mutlu, Phillip B. Gibbons, and Todd C. Mowry, "Ambit: In-Memory Accelerator for Bulk Bitwise Operations Using Commodity DRAM Technology" Proceedings of the 50th International Symposium on Microarchitecture (MICRO), Boston, MA, USA, October 2017.

• Xiangyao Yu, Christopher J. Hughes, Nadathur Satish, Onur Mutlu, and Srinivas Devadas, "Banshee: Bandwidth-Efficient DRAM Caching via Software/Hardware Cooperation" Proceedings of the 50th International Symposium on Microarchitecture (MICRO), Boston, MA, USA, October 2017.

• Samira Khan, Chris Wilkerson, Zhe Wang, Alaa R. Alameldeen, Donghyuk Lee, and Onur Mutlu, "Detecting and Mitigating Data-Dependent DRAM Failures by Exploiting Current Memory Content" Proceedings of the 50th International Symposium on Microarchitecture (MICRO), Boston, MA, USA, October 2017.

• Rachata Ausavarungnirun, Joshua Landgraf, Vance Miller, Saugata Ghose, Jayneel Gandhi, Christopher J. Rossbach, and Onur Mutlu, "Mosaic: A GPU Memory Manager with Application-Transparent Support for Multiple Page Sizes" Proceedings of the 50th International Symposium on Microarchitecture (MICRO), Boston, MA, USA, October 2017.

The following paper has been accepted for publication by the IEEE Transactions on Knowledge and Data Engineering journal.

"High-Level Programming Abstractions for Distributed Graph Processing" by Vasiliki Kalavri, Vladimir Vlassov, and Seif Haridi.

The paper "Error Characterization, Mitigation, and Recovery in Flash-Memory-Based Solid-State Drives" by Yu Cai, Saugata Ghose, Erich F. Haratsch, Yixin Lu and Onur Mutlu has been published in the Proceedings of the IEEE (Volume 105, Issue 9 | September 2017).

Timothy Roscoe gave the talk entitled "Enzian: A Research Computer" at the ARM Research Summit in Cambridge, at MSR Cambridge, and at the University of Cambridge Computer Laboratory.

Onur Mutlu gave an invited talk at the ARM Research Summit on 11 September 2017, entitled "Opportunities and Challenges of Emerging Memory Technologies" .

Slides

Mohsen Ewaida presented the paper "Scalable Inference of Decision Tree Ensembles: Flexible Design for CPU-FPGA Platforms" (Ewaida Mohsen, Hantian Zhang, Ce Zhang and Gustavo Alonso) and David Sidler presented the demo "doppioDB: A Hardware Accelerated Database" (David Sidler, Muhsen Owaida, Zsolt Istvan, Kaan Kara and Gustavo Alonso) at 27th International Conference on Field-Programmable Logic and Applications (FPL 2017), in Ghent, Belgium. 

Merve Gürel joined the Systems Group as a PhD student. Merve completed her Master Thesis at EPFL.

The following five papers and a demo have been presented at the 43rd International Conference on Very Large Data Bases VLDB 2017 in Munich, Germany.

"Distributed Join Algorithms on Thousands of Cores" by Claude Barthels (ETH), Ingo Müller (ETH), Timo Schneider (ETH), Gustavo Alonso (ETH), Torsten Hoefler (ETH)

“Caribou: Intelligent Distributed Storage” by Zsolt Istvan (ETH), David Sidler (ETH), Gustavo Alonso (ETH)

“Fast Scans on Key-Value Store”s by Markus Pilman (ETH), Kevin Bocksrocker (Microsoft), Lucas Braun (ETH), Renato Marroquın (ETH), Donald Kossmann (ETH)

“LDA*: A Robust and Large-scale Topic Modeling System” by Lele Yu (Peking University), Bin Cui (Peking University), Ce Zhang (ETH), Yingxia Shao (Peking University)

“An Experimental Evaluation of SimRank-based Similarity Search Algorithms” by Zhipeng Zhang (Peking University), Yingxia Shao (PKU), Bin Cui (Peking University), Ce Zhang (ETH)

DEMO: “MLog: Towards Declarative In-Database Machine Learning” by Xupeng Li (Peking University), Bin Cui (Peking University), Yiru Chen (Peking University), Wentao Wu (Microsoft Research), Ce Zhang (ETH)

Gustavo Alonso gave a lecture on databases on emerging hardware at the Joint International BBDC and ScaDS Summer School on Big Data in Munich, Germany.

Vasiliki Kalavri gave a turorial at the EIT Big Data Analytics Summer School in Stockholm Sweden. 

Juan Gómez Luna joined the Systems Group as a Postdoc. Juan comes to us from the University of Córdoba.

The following poster has been accepted at the ACM Symposium on Cloud Computing 2017 (SoCC '17), September 25-27, 2017 in Santa Clara, California: 

How Good Are Machine Learning Clouds? by Hantian Zhang, Luyuan Zeng, Wentao Wu (MSR, Redmond), Ce Zhang.

Lukas Arnold joined the Systems Group as a PhD Student. Lukas comes to us from the University of Bristol.

Timothy Roscoe gave the talk "Enzian: a research computer" at Cavium, VMware and Apple in Silicon Valley.

Abstract: 

Traditional systems software research is facing a new challenge to its relevance. Modern hardware CAD systems, the drive to lower power and "dark silicon", FPGAs, and other factors have made it both easy, quick, and cheap for system vendors to build custom hardware platforms. Almost any function can now be put into silicon or reconfigurable logic: the choice of exactly what *should* be built is a short-term business decision based on markets and workloads.

Such hardware qualitatively changes how systems, including system software, should be conceived and designed. However, most published OS research in rack-scale, embedded, or datacenter computing only uses affordable commodity platforms for which documentation is available to researchers. Academia and industry practice are diverging.

Enzian is an attempt to build a computer at ETH Zurich (with help from Cavium and Xilinx) to bridge this gap in a way not possible with commodity hardware or simulation. Enzian nodes closely couple a server-class CPU SoC with a large FPGA in the same coherence domain, with abundant network bandwidth to both chips. It is designed for maximum research flexibility, and can be used in many ways: a high-end server with FPGA-based acceleration, a multiport 200Gb/s NIC supporting custom protocols and cache access, a platform for runtime verification and auditing of system code, to name but three.

Zeke Wang joined the Systems Group as a PostDoc. Zeke comes to us from the National University of Singapore.

Debopam Bhattacherje presented the following paper at HotCloud 2017 in Santa Clara, CA, USA:

"A Cloud-based Content Gathering Network" by Debopam Bhattacherjee, ETH Zurich; Muhammad Tirmazi, LUMS; Ankit Singla, ETH Zurich.

Abstract: 

Many popular Web services use CDNs to host their content closer to users and thus improve page load times. While this model’s success is beyond question, it has its limits: for users with poor last-mile latency even to a nearby CDN node, the many RTTs needed to fetch a Web page add up to large delays. Thus, in this work, we explore a complementary model of speeding up Web page delivery—a content gathering network (CGN), whereby users establish their own geo-distributed presence, and use these points of presence to proxy content for them. We show that deploying only 14 public cloud-based CGN nodes puts the closest node within a median RTT of merely 4.8 ms (7.2 ms) from servers hosting the top 10k (100k) most popular Web sites. The CGN node nearest to a server can thus obtain content from it rapidly, and then transmit it to the client over fewer (limited by available bandwidth) high-latency interactions using aggressive transport protocols. This simple approach reduces the median page load time across 100 popular Web sites by as much as 53%, and can be deployed immediately without depending on any changes to Web servers at an estimated cost of under $1 per month per user.

Vasiliki Kalavri gave the invited tutorial "Programming Models and Tools for Distributed Graph Processing" at the 31st British International Conference on Databases (BICOD) in London.

Abstract: 

Graphs capture relationships between data items, such as interactions or dependencies, and their analysis can reveal valuable insights for machine learning tasks, anomaly detection, clustering, recommendations, social influence analysis, bioinformatics, and other application domains. This tutorial reviews the state of the art in high-level abstractions for distributed graph processing. First, we present six models that were developed specifically for distributed graph processing, namely vertex-centric, scatter-gather, gather-sum-apply-scatter, subgraph-centric, filter-process, and graph traversals. Then, we consider general-purpose distributed programming models that have been used for graph analysis, such as MapReduce, dataflow, linear algebra primitives, datalog, and shared partitioned tables. The tutorial aims at making a qualitative comparison of popular graph programming abstractions. We further consider performance limitations of some graph programming models and we summarize proposed extensions and optimizations.

Reto Achermann gave the talk "Model based system configuration and tasteful hardware" at University of Cambridge.

Yu Liu joined the Systems Group as a 3rd year PhD student.

Vasiliki Kalavri has been awarded the ETH Postdoctoral Fellowship for the project "Automatic scaling of distributed streaming computations using graph analytics on real-time monitoring data".

Torsten Hoefler gave talks, tutorials and participated in panels at ISC 2017 - The event for high performance computing, networking and storage - in Frankfurt, Germany. Details are available here. 

Furthermore, David Sidler presented "doppioDB: A Hardware Accelerated Database" at the Intel Collaboration Hub at ISC 2017. 

Darko Makreshanski defended his PhD Dissertation "Systems and Methods for Interactive Data Processing on Modern Hardware".

The article entitled "Learning Starts with People", published in ETH Globe 2/2017, gives an insight into Ce Zhang's research and his research plans. 

"Fast, intelligent data systems are Ce Zhang’s speciality. To make sure they function smoothly, the data scientist combines basic research with service and dialogue." More...

Vasiliki Kalavri gave the talk “Strymon: Queryable Online Simulation for Modern Datacenters” at the NII Shonan meeting on Language integrated Queries in Kunagawa, Japan.

Timothy Roscoe gave the keynote talk ""The Trouble with Hardware" at the 10th ACM International Systems and Storage Conference (Systor 2017) in Haifa, Israel. 

Ce Zhang gave the talk entitled "Pushing the Boundaries of Machine Learning" at ETH Meets New York.

"The demo: "doppioDB: A Hardware Accelerated Database" by David Sidler, Zsolt Istvan, Muhsen Owaida, Kaan Kara, and Gustavo Alonso was selected as one of the Best Demos of SIGMOD 2017 and won an honorable mention.

The following six papers and a demo have been presented at SIGMOD 2017 and collocated Workshops in Chicago, IL, May 14-19, 2017:

"BatchDB: Efficient Isolated Execution of Hybrid OLTP+OLAP Workloads" by Darko Makreshanski, Jana Giceva, Claude Barthels and Gustavo Alonso, SIGMOD 2017

"Accelerating Pattern Matching Queries in Hybrid CPU-FPGA Architectures" by David Sidler, Zsolt István, Mohsen Ewaida, Gustavo Alonso, SIGMOD 2017

"Heterogeneity-aware Distributed Parameter Servers" by Jiawei Jiang*, Bin Cui*, Ce Zhang, Lele Yu* *(Peking University), SIGMOD 2017

"FPGA Based Data Partitioning" by Kaan Kara, Jana Giceva, Gustavo Alonso, SIGMOD 2017

"An Overreaction to the Broken Machine Learning Abstraction: The ease.ml Vision by Ce Zhang, Wentao Wu (Microsoft Research) and Tian Li (Peking University), Workshop on Human-In-the-Loop Data Analytics (HILDA 2017) collocated with SIGMOD 2017

"Scaling Column Imprints using Advanced Vectorization" by Lefteris Sidirourgos (ETH) and Hannes Muhleisen (CWI), 13th International Workshop on Data Management on New Hardware (DaMoN 2017) collocated with SIGMOD 2017

DEMO: "doppioDB: A Hardware Accelerated Database" by David Sidler, Zsolt Istvan, Muhsen Owaida, Kaan Kara, and Gustavo Alonso, SIGMOD 2017

The following paper has been accepted at the 34th International Conference on Machine Learning (ICML 2017) ICML, Sydney, Australia, August 06-11, 2017: 

The ZipML Framework for Training Models with End-to-End Low Precision: The Cans, the Cannots, and a Little Bit of Deep Learning by Hantian Zhang, Jerry Li, Kaan Kara, Dan Alistarh, Ji Liu, Ce Zhan.

The following two papers have been presented at the 25th IEEE International Symposium on Field-Programmable Custom Computing Machines (FCCM 2017) in Napa, CA, US, April 30 - May 2, 2017:

"Centaur: A Framework for Hybrid CPU-FPGA Databases" by Muhsen Owaida, David Sidler, Kaan Kara and Gustavo Alonso

"FPGA accelerated Dense Linear Machine Learning: A Precision-Convergence Trade-off" by Kaan Kara, Dan Alistarh, Ce Zhang, Onur Mutlu, Gustavo Alonso

The following demo co-authored by Ce Zhang has been accepted at VLDB 2017 Demonstration Track, Munich, Germany, August 28 to September 1, 2017.

MLog: Towards Declarative In-Database Machine Learning by Xupeng Li*, Bin Cui*, Yiru Chen*, Wentao Wu**, Ce Zhang.

*Peking University, China
**MSR Redmond, USA

John Liagouris gave the talk entitled "Understanding Distributed Dataflow Systems" at Google, Mountain View and at VMware, Palo Alto.

The paper "A Cloud-based Content Gathering Network" by Debopam Bhattacherjee (ETH Zurich), Muhammad Tirmazi (LUMS), Ankit Singla (ETH Zurich) has been accepted at HotCloud 2017, Santa Clara, CA, July 10-11, 2017.

Zaheer Chothia presented the following paper at EuroSys 2017 in Belgrade, Serbia:

Online Reconstruction of Structural Information from Datacenter Logs by Zaheer Chothia, John Liagouris, Desislava Dimitrova, and Timothy Roscoe.

Abstract:

Well-run datacenter application architectures are heavily instrumented to provide detailed traces of messages and remote invocations. Reconstructing user sessions, call graphs, transaction trees, and other structural information from these messages, a process known as sessionization, is the foundation for a variety of diagnostic, profiling, and monitoring tasks essential to the operation of the datacenter.

We present the design and implementation of a system which processes log streams at gigabits per second and reconstructs user sessions comprising millions of transactions per second in real time with modest compute resources, while dealing with clock skew, message loss, and other real-world phenomena that make such a task challenging. Our system is based on the Timely Dataflow framework for low latency, data-parallel computation, and we demonstrate its utility with a number of use-cases and traces from a large, operational, mission-critical enterprise data center.

The following paper has been accepted at HotOS XVI, Whistler, British Columbia, Canada, May 7-10, 2017.

Separating Translation from Protection in Address Spaces with Dynamic Remapping by Reto Achermann, Chris Dalton (HP Labs), Paolo Faraboschi (HP Labs), Moritz Hoffman, Dejan Milojicic (HP Labs), Geoffrey Ndu (HP Labs), Alexander Richardson (University of Cambridge), Timothy Roscoe, Adrian L. Shaw (HP Labs), Robert N. M. Watson (University of Cambridge).

The paper "An Overreaction to the Broken Machine Learning Abstraction: The ease.ml Vision by Ce Zhang, Wentao Wu (Microsoft Research) and Tian Li (Peking University) has been accepted at the Workshop on Human-In-the-Loop Data Analytics (HILDA 2017) co-located with SIGMOD 2017 (14 May 2017, Chicago).

The following paper has been accepted at 13th International Workshop on Data Management on New Hardware (DaMoN 2017) collocated with 2017 ACM SIGMOD/PODS Conference, Chicago, IL USA, May 14th - 19th, 2017:

Scaling Column Imprints using Advanced Vectorization by Lefteris Sidirourgos (ETH) and Hannes Muhleisen (CWI).

Timothy Roscoe gave the talk entitled "Diagnosing Datacenters" at Amadeus Architecture, Quality & Governance.

The paper entitled "Improving the Reliabililty of Chip-off Forensic Analysis of NAND Flash Memory Devices" by Aya Fukami (Japan National Police Abgency); Saugata Ghose, Yixin Luo, Yu Cai (Carnegie Mellon University); and Onur Mutlu (ETH Zurich) has been selected as the BEST PAPER at the 4th European Digital Forensics Research Workshop and Conference (DFRWS 2017) held in Überlingen, Lake Constance, Germany.

The following paper has been accepted at EuroSys 2017 in Belgrade, Serbia, April 23-26, 2017:

Online Reconstruction of Structural Information from Datacenter Logs by Zaheer Chothia, John Liagouris, Desislava Dimitrova, and Timothy Roscoe.

The following demo has been accepted at SIGMOD 2017, Chicago, IL, May 14-19, 2017: "doppioDB: A Hardware Accelerated Database" by David Sidler, Zsolt Istvan, Muhsen Owaida, Kaan Kara, and Gustavo Alonso.

Ce Zhang's collaboration with ETH astrophysicist, Kevin Shawinski, has been followed by the media since the paper "Generative Adversarial Networks recover features in astrophysical images of galaxies beyond the deconvolution limit" by Kevin Schawinski, Ce Zhang, Hantian Zhang, Lucas Fowler, and Gokula Krishnan has recently been published in Monthly Notices of the Royal Astronomical Society.

The article "Machine Learning Is Bringing the Cosmos Into Focus" has been published in "The Atlantic".

More press coverage:

Oxford Journals: https://oxfordjournals.altmetric.com/details/16663618/news

Phys.org: https://phys.org/news/2017-02-neural-networks-sharpest-images.html 

Project page: space.ml

Markus Pilman defended his PhD Dissertation "Tell: An Elastic Database System for Mixed Workloads".

Tal Ben Nun joined the Systems Group as Postdoc on 20 February 2017. Tal comes to us from the Hebrew University of Jerusalem.

Andrea Lattuada joined the Systems Group as PhD student. Andrea completed his Master Thesis in the Systems Group.

Lois Orosa joined the Systems Group as Academic Guest. Lois comes to us from the University of Campinas in Brazil.

Minesh Patel und Jeremie Kim joined the Systems Group as PhD students. Minesh comes to us from the University of Texas, Austin and Jeremie from Carnegie Mellon University, Pittsburg.

The following paper has been accepted at SIGMOD 2017 in Raleigh, NC, USA; May 14-19, 2017.

BatchDB: Efficient Isolated Execution of Hybrid OLTP+OLAP Workloads by Darko Makreshanski, Jana Giceva, Claude Barthels and Gustavo Alonso.

Jana Giceva defended her PhD Dissertation "Database/Operating system co-design".

Hasan Hassan joined the Systems Group as a new PhD student. He completed his master thesis at TOBB University of Economics and Technology, Ankara, Turkey.

Besmira Nushi presented the paper "A Human-in-the-loop Approach for Troubleshooting Machine Learning Systems"(Besmira Nushi, Ece Kamar, Donald Kossmann and Eric Horvitz) at the Future of Interactive Learning Machines Workshop (FILM) collocated with NIPS 2016 in Barcelona, Spain.

Ce Zhang gave the talk "Machine learning on the edge and beyond" at Schindler in Luzern.

The following four papers have been presented at The International Conference for High Performance Computing, Networking, Storage and Analysis (SC'16) in Salt Lake City:

Scheduling-Aware Routing for Supercomputers by J. Domke and T. Hoefler

dCUDA: Hardware Supported Overlap of Computation and Communication by T. Gysi, J. Baer, T. Hoefler

Extreme Scale Plasma Turbulence Simulations on Top Supercomputers Worldwide by W. Tang, B. Wang, S. Ethier, G. Kwasniewski, T. Hoefler, K. Ibrahim, K. Madduri, S. Williams, L. Oliker, C. Rosales-Fernandez, T. Williams

A PCIe Congestion-Aware Performance Model for Densely Populated Accelerator Servers by M. Martinasso, G. Kwasniewski, S. Alam, T. Schulthess, T. Hoefler.

Gustavo Alonso gave the talks "Data Processing on the Fast Lane" at Intel Santa Clara and at the Workshop on Reconfigurable Hardware for HPC held jointly with Supercomputing. He also gave the talk "Accelerating Database Engines with FPGAs" at Xilinx.

The paper "On Human Intellect and Machine Failures: Troubleshooting Integrative Machine Learning Systems"  by Besmira Nushi, Ece Kamar, Eric Horvitz, Donald Kossmann has been accepted at The Thirty-First AAAI Conference on Artificial Intelligence (AAAI-17), which will be held February 4-9, 2017 in San Francisco.

Ce Zhang gave the talk entitled "Accessible Data Sciences with Efficient Data Systems" at IBM Zurich Research Lab.

Abstract: 

One important problem for the current state of data science is that many of the techniques needed to unleash the next big thing are available but still far from accessible. Specifically, the current machine-learning ecosystems are difficult to use by non-computer science users and they are still far from achieving the full potential that can be provided by modern hardware. With more than five ongoing data sciences applications here at ETH Zurich, ranging from genomics, social sciences, and astronomy, our dream is to design the next generation of data science ecosystems that are fast, scalable, and easier to use. In this talk, I will first describe the abundant opportunities for data sciences at ETH Zurich, and then describe two enabling techniques that are being developed by my group. The general direction of these techniques is the co-design of machine learning (or artificial intelligence) with modern hardware and systems. I will talk about our recent work that introduced a data structure for dense linear regression. It can potentially reduce the memory bandwidth by 20x while training. Then I will introduce a novel database architecture, which makes the production system of a leading security company 100x faster. It contains an SMT solver to answer queries that it was not originally designed for.

Andrei Marian Dan, Patrick Lam, Torsten Hoefler, and Martin Vechev won the Distinguished Paper Award at The International Conference on Object-Oriented Programming, Systems, Languages, and Applications (OOPSLA 2016) for their paper “Modeling and Analysis of Remote Memory Access Programming”. 

The paper entitled "MQJoin: Efficient Shared Execution of Main-memory Joins" by Darko Makreshanski, Georgios Giannikis, Gustavo Alonso and Donald Kossmann has been selected as one of the Best Papers of VLDB 2016.

Abstract: 

Database architectures typically process queries one-at-a-time, ex- ecuting concurrent queries in independent execution contexts. Of- ten, such a design leads to unpredictable performance and poor scalability. One approach to circumvent the problem is to take advantage of sharing opportunities across concurrently running queries. In this paper we propose Many-Query Join (MQJoin), a novel method for sharing the execution of a join that can effi- ciently deal with hundreds of concurrent queries.

This is achieved by minimizing redundant work and making efficient use of main- memory bandwidth and multi-core architectures. Compared to ex- isting proposals, MQJoin is able to efficiently handle larger work- loads regardless of the schema by exploiting more sharing oppor- tunities. We also compared MQJoin to two commercial main- memory column-store databases. For a TPC-H based workload, we show that MQJoin provides 2-5x higher throughput with signif- icantly more stable response times.

Onur Mutlu gave the talk entitled "Rethinking Memory System Design (and the Computing Platforms We Design Around It)" at IBM Research Zurich.

Abstract:

The memory system is a fundamental performance and energy bottleneck in almost all computing systems. Recent system design, application, and technology trends that require more capacity, bandwidth, efficiency, and predictability out of the memory system make it an even more important system bottleneck. At the same time, DRAM and flash technologies are experiencing difficult technology scaling challenges that make the maintenance and enhancement of their capacity, energy efficiency, and reliability significantly more costly with conventional techniques. In fact, recent reliability issues with DRAM, such as the RowHammer problem, are already threatening system security and predictability.

In this talk, we first discuss major challenges facing modern memory systems in the presence of greatly increasing demand for data and its fast analysis. We then examine some promising research and design directions to overcome these challenges and thus enable scalable memory systems for the future. We discuss three key solution directions: 1) enabling new memory architectures, functions, interfaces, and better integration of memory and the rest of the system, 2) designing a memory system that intelligently employs emerging non-volatile memory (NVM) technologies and coordinates memory and storage management, 3) reducing memory interference and providing predictable performance to applications sharing the memory system. If time permits, we will also touch upon our ongoing related work in combating scaling challenges of NAND flash memory.

An accompanying paper, slightly outdated (circa 2015), can be found here.

Yaohua Wang, Postdoc from the National Univ. of Defense Technology in Changsha, China, joined the Systems Group as Academic Guest. Yoahua will be staying with us until October 2017.

John Liagouris gave the talk “Explaining Outputs in Modern Data Analytics” at the Department of Computer Science, University of Hong Kong (HKU).

Ce Zhang gave the talk entitled "Accessible Data Sciences with Efficient Data Systems" at following universities in China:

• Tsinghua University, Beijing
• Peking Univeristy, Beijing
• National University of Defense Technology, Changsha
• Fudan University, Shanghai • Harbin Institute of Technology, Harbin
• East China Normal University, Shanghai
• Shanghai Jiaotong University, Shanghai

Jana Giceva gives the talk entitled "Customized OS support for data processing on modern hardware" at the SFB Colloquium (TU Dortmund).

The paper entitled "READY: Completeness is in the Eye of the Beholder" by Badrish Chandramouli, Johannes Gehrke, Jonathan Goldstein, Donald Kossmann, Justin Levandoski, Renato Marroquin and Wenlei Xie has been accepted at CIDR 2017.

Ce Zhang gave the talk entitled "Data Science Applications and Machine Learning/System Co-Design" at Huawei European Research Center, Munich, Germany.

The paper entitled "A Survey of Microarchitectural Timing Attacks and Countermeasures on Contemporary Hardware" by Qian Ge, Yuval Yarom, David Cock and Gernot Heiser, will be published in the Journal of Cryptographic Engineering (JCEN).

Abstract: Microarchitectural timing channels expose hidden hardware state though timing. We survey recent attacks that exploit microarchitectural features in shared hardware, especially as they are relevant for cloud computing. We classify types of attacks according to a taxonomy of the shared resources leveraged for such attacks. Moreover, we take a detailed look at attacks used against shared caches. We survey existing countermeasures. We finally discuss trends in the attacks, challenges to combating them, and future directions, especially with respect to hardware support.

Onur Mutlu gave the keynote talk "Rethinking Memory System Design" at 27th International Symposium on Rapid System Prototyping (RSP 2016) in Pittsburgh, PA, USA.

Abstract:

The memory system is a fundamental performance and energy bottleneck in almost all computing systems. Recent system design, application, and technology trends that require more capacity, bandwidth, efficiency, and predictability out of the memory system make it an even more important system bottleneck. At the same time, DRAM and flash technologies are experiencing difficult technology scaling challenges that make the maintenance and enhancement of their capacity, energy efficiency, and reliability significantly more costly with conventional techniques. In fact, recent reliability issues with DRAM, such as the RowHammer problem, are already threatening system security, predictability and robustness.

In this talk, we first discuss major challenges modern memory systems face in the presence of increasing demand for data and its fast analysis. We then examine some promising research and design directions to overcome these challenges and enable scalable memory systems for the future. We discuss three key solution directions: 1) enabling new memory architectures, functions, interfaces, and better integration of memory and the rest of the system, 2) designing a memory system that intelligently employs emerging non-volatile memory (NVM) technologies, 3) reducing memory interference and providing predictable performance to applications sharing the memory system. If time permits, we will also touch upon our ongoing related work in combating scaling challenges of NAND flash memory.

The Interview with Torsten Hoefler entitled "How to program large-scale heterogeneous parallel computers" has been published as Spotlight on the ETH Department of Computer Science Web page.

Tobias Grosser received an SNSF Ambizione Fellowship to work the next three years on heterogeneous compute in http://polly.llvm.org/.

The paper "Fat-Free Topologies" by Ankit Singla has been accepted at HotNets 2016.

Anja Grünheid defended her PhD Dissertation entitled “Data Integration with Dynamic Data Sources”.

Anja took up a position with Google, New York.

Timothy Roscoe gave the talk entitled “Barrelfish: an OS for real, modern hardware” at Cavium in San Jose. At Google in Mountain View he gave the talk entitled “Online simulation as a diagnostic tool for datacenters".

The following two Systems Group papers have been presented at VLDB 2016 in New Delhi:

Darko Makreshanski presented: "MQJoin: Efficient Shared Execution of Main-Memory Joins" by Darko Makreshanski, Georgios Giannikis, Gustavo Alonso and Donald Kossmann.

John Liagouris presented: "Explaining Outputs in Modern Data Analytics" by Zaheer Chothia, John Liagouris, Frank McSherry and Timothy Roscoe.

The following two short papers have been presented at 26th International Conference on Field-Programmable Logic and Applications (FPL'16) in Lausanne:

David Sidler presented: "Low-Latency TCP/IP Stack for Data Center Applications" by David Sidler, Zsolt István and Gustavo Alonso.

Kaan Kara presented: "Fast and Robust Hashing for Database Operators" by Kaan Kara and Gustavo Alonso.

Ce Zhang has joined the Systems Group as Assistant Professor. 

Zhang's research interests are databases, data processing, machine learning and data science. His applications focus on systems to help scientists analyse and understand large quantities of data. He links conventional areas of databases and information retrieval with new methods of machine learning. More...

Timo Schneider received the Best Student Paper Award at Hot Interconnects 2016 for the paper "Ensuring Deadlock-Freedom in Low-Diameter InfiniBand Networks" by Timo Schneider, Otto Bibartiu, Toersten Hoefler.

Vojislav Dukic joined the Systems Group as PhD Student on 15 August 2016.

Timothy Roscoe interviewed in "The Register" on 29 July 2016

"The bigger they get, the harder we fall: Thinking our way out of cloud crash"

Ce Zhang, from Stanford University, will join the Computer Science Department and the Systems Group as assistant professor in September 2016.

The paper entitled "Measuring and Understanding Throughput of Network Topologies" by Sangeetha Abdu Jyothi, Ankit Singla, P. Brighten Godfrey, and Alexandra Kolla has been accepted at Super Computing 2016 in Salt Lake City, Utah, USA, 13-18 Nov. 2016.

David Sidler und Kaan Kara are attending the eleventh Microsoft Research Summer School at Microsoft Research in Cambridge, UK.

Gustavo Alonso talks about the collaboration with Microsoft Research - Swiss Joint Research Center in an interview published as a Departmental Spotlight.

The paper "Explaining Outputs in Modern Data Analytics" by Zaheer Chothia, John Liagouris, Frank McSherry and Timothy Roscoe has been accepted for presentation at VLDB 2016 in New Delhi, India, Sept. 5 - 9, 2016.

Timothy Roscoe gives a talk entitled "Barrelfish: an operating system for modern hardware" at USENIX Vail Computer Elements Workshop (VCEW 2016) in Vail, CO, USA.

Ankit Singla talks about his plans at ETH in the "Spotlight" interview published on the ETH Department of Computer Science Web page.

Gustavo Alonso gave a talk entitled "Accelerating Data Science" at the 2nd Portugal|UT Austin Summer School in Systems and Networking in Costa da Caparica, Lisbon, Portugal.

Moritz Hoffmann joined HPE Labs in Palo Alto, CA, USA for an Internship until September 16th, 2016.

Onur Mutlu joined the Systems Group as a Full Professor.

Onur conducts research in in computer architecture, systems, and bioinformatics. His work spans and stretches the boundaries between applications, systems, languages, system software, compilers, and hardware. His research tackles many issues in high performance, energy efficiency, hardware security, fault tolerance, predictable systems, dependable systems, and hardware/software cooperation. He is especially excited about novel computation, communication and memory/storage paradigms, applied to emerging systems, technologies, and bioinformatics/medical applications. He is also excited about system design for bioinformatics and biologically inspired computing paradigms.

David Sidler presents the paper entitled "Runtime Parameterizable Regular Expression Operators for Databases" by Zsolt Istvan, David Sidler and Gustavo Alonso at FCCM 2016 in Washington DC on May 2, 2016.

Besmira Nushi gave a talk entitled “Crowdsourcing on a budget: Plan early, plan twice!” at Microsoft Research Redmond. She presented her work on learning and making predictions from crowdsourced data under budget constraints.

Ankit Singla talks about the Internet at the speed of light (Ein Lichtgeschwindigkeits-Internet) in Electro Suisse Bulletin.

The Systems Group is presenting at two Workshops at EuroSys 2016 in London, April 18-21, 2016:

Jana Giceva: Basslet: an OS Runtime for Parallel Data Processing (Jana Giceva, Gerd Zellweger, Gustavo Alonso and Timothy Roscoe) - Workshop on Multicore and Rack-scale Systems (MaRS'16)

David Cock: Litmus Testing at Rack Scale  - Workshop on Multicore and Rack-scale Systems (MaRS'16)

Roni Häcki: Replication in Rack-scale Systems  -  10th EuroSys Doctoral Workshop  (EuroDW'16)

Gustavo Alonso gave a keynote talk entitled "Generalization versus Specialization in cloud computing infrastructures" at IEEE International Conference on Cloud Engineering (IC2E) 2016 in Berlin. 

Abstract :

Cloud computing represents a fundamental change in the business model behind IT: a shift from manufacturing of software and hardware products towards packaging infrastructure, processing, and storage as services. Cloud data centers, given their intended use for general purpose computing, would seem to push towards homogeneity in architectures and platforms. Modern applications and use cases, from scientific computing to big data, push in exactly the opposite direction: an increase in specialization as a way to efficiently meet demanding requirements. In this talk I will illustrate both trends and argue that, contradictory as they seem to be, there are many opportunities in combining them. Doing so requires to work in two areas. One is to find better ways to extend the performance and efficiency advantages of specialization to general purpose settings. The other is to develop the necessary software and hardware layers to allow generalized use of specialized systems. Taking together, these efforts outline an exciting research and development landscape that I will outline as a conclusion of the talk.

Zsolt István presented the following paper at NSDI 2016 in Santa Clara, CA:

Consensus in a Box: Inexpensive Coordination in Hardware by Zsolt István, David Sidler, and Gustavo Alonso, ETH Zürich; Marko Vukolić, IBM Research—Zürich.

David Sidler gave the following talk at Oracle Labs (Redwood City, CA) and at Intel Labs (Santa Clara, CA): "Accelerating String Matching Queries with Hybrid CPU-FPGA Multicores".

Timothy Roscoe gives the following talk at VMware in Palo Alto, CA, USA:

Title: "Physical Memory Management for Modern Hardware"

Abstract:

Classical VM is an opaque abstraction of RAM, backed by demand paging. However, most systems today (from phones to data-centers) do not page, and indeed may require the performance benefits of non-paged physical memory, precise NUMA allocation, and distinguish very different kinds of memory (such as NVRAM). Moreover, MMU hardware is now useful for other purposes, such as detecting page access or providing large page translation. Accordingly, the venerable VM abstraction in OSes like Windows and Linux has acquired a plethora of extra APIs to poke at the policy behind the illusion of a virtual address space.

In the Barrelfish OS, we rethink how an OS supports virtual memory. The Barrelfish memory system which inverts the traditional model:. Applications explicitly manage their physical RAM of different types, and directly (though safely) program the translation hardware. The approach requires no virtualization support, and outperforms VMM-based approaches for all but the smallest working sets. I'll talk about use-cases for virtual memory not possible in Linux-like systems today, and show that other use-cases are simple to program and significantly faster.

Lukas Mosimann joined the Systems Group as a new PhD student by Prof. Torsten Hoefler.

Gustavo Alonso gave a talk entitled "Databases in the Cloud: they have to be different" at the "Oracle Labs Workshop on Databases in the Cloud" (TU Munich, Germany), Feb 10-11, 2016.

Lukas Humbel joined the Systems Group as a PhD student. Lukas completed his master thesis in the Systems Group.

The following demo has been accepted at ICDE 2016, Helsinki, Finland, May 16-20, 2016:

"graphVizdb: A Scalable Platform for Interactive Large Graph Visualization" by Nikos Bikakis (NTU Athens), John Liagouris (ETH Zürich), Maria Krommyda (NTU Athens), George Papastefanatos (IMIS, Research Center 'Athena'), Timos Sellis (RMIT University).

Ankit Singla joined the Systems Group as an Assistant Professor. More information about Ankit Singla can be found here.

.

We have welcomed two new Systems Group members today:

Adam Turowski - Software Engineer

Ingo Müller - PostDoc

Torsten Hoefler's  project about weather and climate models has been broadcasted at BBC Arabic.

The paper entitled "Zookeeper in a Box: Hardened Consensus for Inexpensive Coordination" by Zsolt István, David Sidler, Gustavo Alonso and Marko Vukolic (IBM Zurich) has been accepted at NSDI 2016, Santa Clara, CA, USA, March 16-18, 2016.

Our alumnus, Michael Duller, joined Datometry, a start-up company in San Francisco. 

The following paper has been accepted at ASPLOS 2016 in Atlanta, Georgia, USA, April 2-6, 2016: 

"SpaceJMP: Programming with Multiple Virtual Address Spaces" by Izzat El Hajj (University of Illinois at Champaign-Urbana) Alexander Merritt (Georgia Institute of Technology) Gerd Zellweger (ETH Zürich) Dejan Milojicic (HP Labs) Wen-Mei Hwu (University of Illinois at Champaign-Urbana) Karsten Schwan (Georgia Institute of Technology) Timothy Roscoe (ETH Zürich) Reto Achermann (ETH Zürich) Paolo Faraboschi (HP Labs)

Besmira Nushi presented the paper "Crowd Access Path Optimization: Diversity Matters" (Besmira Nushi, Adish Singla, Anja Gruenheid, Erfan Zamanian, Andreas Krause, Donald Kossmann) at the Conference on Human Computation & Crowdsourcing (HCOMP 2015), in Kona Kai Resort and Marina, San Diego, USA.

The article entitled "Computing for Climate (part 1): Evolution of Models" by Torsten Hoefler, Christoph Schär (ETH Zurich) and Oliver Fuhrer (MeteoSchweiz) was published in ETH Zukunftsblog. 

Nadia Mouci joined the Systems Group today as our new administrative assistant.

The following Systems Group papers have been presented at VLDB 2015 in Kohala Coast, Hawai'I, VLDB 2015, August 31-Sept. 4, 2015:

• "Deployment of query plans on multicores" by Jana Giceva, Gustavo Alonso, Timothy Roscoe and Tim Harris (Oracle Labs Cambridge)
• "To Lock, CAS, or Elide: On the Interplay of Hardware Transactional Memory and Lock-Free Indexing" by Darko Makreshanski, Justin Levandoski (MSR), and Ryan Stutsman (MSR)
• "Indexing and Selecting Hierarchical Business Logic" by Alessandra Loro, Anja Gruenheid, Donald Kossmann, Damien Profeta (Amadeus) & Philippe Beaudequin (Amadeus)
• "On Uncertain Graphs Modeling and Queries" Arijit Khan and Lei Chen (Hong Kong University of Science and Technology) - tutorial

Daniela Dorneanu took up a Solution Developer and Trainer position at Appway, Zürich.

The following PhD students left for their internships this summer:

Gerd Zellweger – HP, Palo Alto, CA, USA

Anja Grünheid – Microsoft Research, Redmond, WA, USA

Renato Marroquin – NASA Jet Propulsion Laboratory, Pasadena, CA, USA

David Sidler – Microsoft Research, Redmond, WA, USA

Besmira Nushi – Microsoft Research, Redmond, WA, USA

The paper entitled "Top-k Reliable Edge Colors in Uncertain Graphs" by Arijit Khan, Francesco Gullo (Yahoo! Labs Barcelona), Thomas Wohler (ETH), and Francesco Bonchi (Yahoo! Labs Barcelona) has been accepted as a short paper at the CIKM 2015 in Melbourne, Australia, Oct 19-23, 2015.

Frank McSherry joined the Systems Group today as our new Senior Scientist. His research will be in the area of Big Data system design and implementation.

The following tutorial by Arijit Khan and Lei Chen (Hong Kong University of Science and Technology)  has been accepted at VLDB 2015,  August 31-Sept. 4, 2015, Kohala Coast, Hawai'i.

Arijit Khan will join the School of Computer Engineering (SCE) at NTU Singapore as a tenure-track assistant professor starting December 22, 2015.

Zsolt István gives the talk entitled "Better Energy Efficiency and Performance in Datacenters through Specialization" at Microsoft Research (Redmond, USA) and at University of Washington.

Donald Kossman gave a keynote talk entitled “Cipherbase: End-to-end Encryption for SQL” at WAIM 2015 in Qingdao, Shandong, China.

Abstract:

This talk presents the design of the Cipherbase system. Cipherbase was designed to protect "data-in-use". That is, data remains encrypted even when it is processed. To this end, Cipherbase features a novel fine-grained architecture that farms out all computations on encrypted data to an FPGA. The talk provides the motivation and design space for systems like Cipherbase and presents the results of recent performance experiments with the Cipherbase prototype developed at Microsoft Research.

The following four papers and a demo have been presented at SIGMOD 2015, in Melbourne, Australia.

Claude Barthels, Simon Loesing, Donald Kossmann, Gustavo Alonso. Rack-Scale InMemory Join Processing using RDMA. SIGMOD 2015.

Lucas Braun, Thomas Etter, Georgios Gasparis, Martin Kaufmann, Donald Kossmann, Daniel Widmer. Analytics in Motion: High Performance Event-Processing AND Real-Time Analytics in the Same Database. SIGMOD 2015.

Phil Bernstein, Sudipto Das, Bailu Ding, Markus Pilman. Optimizing Optimistic Concurrency Control for Tree-Structured, Log-Structured Databases. SIGMOD 2015.

Simon Loesing, Markus Pilman, Thomas Etter, Donald Kossmann. On the Design and Scalability of Distributed Shared-Data Databases. SIGMOD 2015.

Anja Gruenheid, Theodoros Rekatsinas, Donald Kossmann, Divesh Srivastava. DEMO: StoryPivot: Comparing and Contrasting Story Evolution. SIGMOD 2015.

The paper entitled "To Lock, CAS, or Elide: On the Interplay of Hardware Transactional Memory and Lock-Free Indexing" by Darko Makreshanski, Justin Levandoski (MSR), and Ryan Stutsman (MSR) has been accepted at VLDB 2015, August 31-Sept. 4, 2015, Kohala Coast, Hawai'i.

David Sidler presented his work on TCP/IP for FPGAs at FCCM 2015, in Vancouver, Canada.

"Scalable 10 Gbps TCP/IP Stack Architecture for Reconfigurable Hardware" by David Sidler, Michaela Blott, Kimon Karras, Raymond Carley, Gustavo Alonso and Kees Vissers.

The following paper has been accepted at the 2015 USENIX Annual Technical Conference (USENIX ATC), Santa Clara, CA, USA; July 8-10, 2015.

Shoal: smart allocation and replication of memory for parallel programs by Stefan Kaestle, Reto Achermann, Timothy Roscoe (Systems Group, Dept. of Computer Science, ETH Zurich), Tim Harris (Oracle Labs, Cambridge, UK)

Jana Giceva has presented the FruitBox project at the Rack Scale Computing Workshop at EuroSys 2015, Bordeaux, France.

The following paper has been accepted for publication in the IEEE Transactions on Knowledge and Data Engineering (TKDE) Journal:

"Querying Knowledge Graphs by Example Entity Tuples". Nandish Jayaram (UT Arlington), Arijit Khan, Chengkai Li (UT Arlington), Xifeng Yan (UCSB), and Ramez Elmasri (UT Arlington)

Selected as one of the top papers of FPL'13, an extended version of the work on hash table design on FPGAs by Zsolt Istvan has appeared in ACM TRETS:

A Hash Table for Line-Rate Data Processing, Z. István, G. Alonso, M. Blott, K.Vissers ACM Transactions on Reconfigurable Technology and Systems (TRETS) 8 (2), March 2015

This year four papers and one demo have been accepted at SIGMOD 2015 in Melbourne, Australia, May 21 - June 4, 2015: 

Claude Barthels, Simon Loesing, Donald Kossmann, Gustavo Alonso. Rack-Scale InMemory Join Processing using RDMA. SIGMOD 2015.

Lucas Braun, Thomas Etter, Georgios Gasparis, Martin Kaufmann, Donald Kossmann, Daniel Widmer. Analytics in Motion: High Performance Event-Processing AND Real-Time Analytics in the Same Database. SIGMOD 2015.

Phil Bernstein, Sudipto Das, Bailu Ding, Markus Pilman. Optimizing Optimistic Concurrency Control for Tree-Structured, Log-Structured Databases. SIGMOD 2015.

Simon Loesing, Markus Pilman, Thomas Etter, Donald Kossmann. On the Design and Scalability of Distributed Shared-Data Databases. SIGMOD 2015.

Anja Gruenheid, Theodoros Rekatsinas, Donald Kossmann, Divesh Srivastava. DEMO: StoryPivot: Comparing and Contrasting Story Evolution. SIGMOD 2015.

The following Workshop paper and 3 posters have been accepted at EUROSYS 2015.

Jana Giceva, Darko Makreshanski, Claude Barthels, Alessandro Dovis, Gustavo Alonso, Donald Kossmann. Rack - scale Data Processing System. WRSC Workshop on Rack - Scale Computing.

Stefan Kaestle, Reto Achermann, Timothy Roscoe. Shoal: smart allocation and replication of memory for parallel programs.- Poster

Pravin Shinde, Kornilios Kourtis , Antoine Kaufmann (University of Washington) and Timothy Roscoe. Intelligent NIC Queue Management in the Dragonet Network Stack - Poster

Moritz Hoffmann: Rack - aware operating systems. Doctoral Workshop. - Poster

John Liagouris joined the System Group as a new Postdoc. John is coming to us from the National Technical University of Athens (NTUA).

Daniela Dorneanu gave a talk entitled "CAPI/Flash + PackedObjects, a novel off-heap solution for scaling-up complex Java applications" at IBM Research Rüschlikon. In this talk Daniela presented her work done at IBM Research (Austin, Texas) during her internship last year.

Jonas Pfefferle joins the Systems Group as new PhD student.

The Workshop proposal: Big-O(Q): Big Graphs Online Querying, organized by Arijit Khan (System Group, ETH), Prasenjit Mitra (Penn State University), Cong Yu (Google Research, New York) has been accepted at VLDB 2015.

Martin Kaufmann joins Deloitte in Zürich as a Consultant.

David Cock joined the Systems Group as a new Past Doc. David comes to us from NICTA (National ICT Australia).

Simon Loesing will join ELCA Zürich as a Software Architect in March 2015.

Tahmineh Sanamrad defended her PhD Dissertation entitled: Encrypting Databases in the Cloud: Threats and Solutions

Committee:

• Prof. Donald Kossman
• Prof. Gustavo Alonso
• Prof. Srdjan Capkun
• Prof. Johannes Gehrke (Cornell University)

Tahmine will join Google Zürich in January 2015.

Moritz Hoffmann joined the Systems Group as a new PhD student. Moritz completed his Master Thesis in our group.

Louis Woods has been awarded the 2015 ETH Medal for his dissertation: FPGA-Enhanced Data Processing Systems.

The paper entitled "Arrakis: The Operating System is the Control Plane" by Simon Peter (Systems Group alumnus), Jialin Li, Irene Zhang, Dan R. K. Ports, Doug Woos, Arvind Krishnamurthy, Thomas Anderson, and Timothy Roscoe won the "Jay Lepreau Best Paper" Award at OSDI 2014 in Broomfield, Colorado.

In addition, one of the other two best papers was the paper with our alumnus Andrew Baumann as the first author: 

Shielding Applications from an Untrusted Cloud with Haven by Andrew Baumann, Marcus Peinado, and Galen Hunt (Microsoft Research)

Marko Vukolic, faculty at EUROCOM in France, has joined our group as a visiting professor. Marko works in distributed computing and will stay with us for the next months.

Cagri Balkesen has obtained an Excellent Presentation Award at VLDB 2014 for his talk on implementing hash and sort-merge joins.

Daniela will be doing her Internship with IBM Research in Austin, Texas from September to December 2014.

In addition to the paper on Pydron, the two following papers were accepted at OSDI 2014:

"Decoupling Cores, Kernels, and Operating Systems" by Gerd Zellweger, Simon Gerber, Kornilios Kourtis, and Timothy Roscoe

"Arrakis: The Operating System is the Control Plane" by Simon Peter, Jialin Li, Irene Zhang, Dan R. K. Ports, Doug Woos, Arvind Krishnamurthy, Thomas Anderson, and Timothy Roscoe

The following paper was presented by Tahmineh Sanamrad at the Conference on Data and Application Security and Privacy (DBSEC 2014) in Vienna, July 14-16, 2014:

“Randomly Partitioned Encryption on Cloud Databases” by Tahmineh Sanamrad, Lucas Braun, Donald Kossmann and Ramarathnam Venkatesan (Microsoft Research)

Qin Yin joined Google Zürich today as a software engineer.

System Group alumna Angela Nicoara has joined Intel Labs in Portland, Oregon.

The following Systems Group papers are being presented at SIGMOD 2014 in Snowbird, Utah, USA, June 22-27, 2014:

• "Histograms as a Side Effect of Data Movement for Big Data" by Zsolt Istvan, Louis Woods, Gustavo Alonso. (presented by Zsolt Istvan)
• “Towards Indexing Functions: Answering Scalar Product Queries” by Arijit Khan, Pouya Yanki, Bojana Dimcheva, and Donald Kossmann. (presented by Arijit Khan)

At the Big Uncertain Data Workshop (BUDA) Workshop, affiliated with SIGMOD 2014, Anja Grünheid is presenting the following paper:

• “Cost-Efficient Querying Strategies for the Crowd" by Anja Grünheid, Besmira Nushi, and Donald Kossmann.

Timothy Roscoe, Tom Anderson, Mic Bowman, David Culler and Larry Peterson, received the 2014 Software Tools User Group (STUG) Award for their work on the Planet Lab Software System. 

The impact of the Planet Lab Software System has been recently recognized through the following two "Test of Time" awards: 

The paper "Operating system support for planetary-scale network services" co-authored by Timothy Roscoe received the first NSDI 10-year "Test of Time" award. 

Timothy Roscoe was one of the recipients of the 2013 ACM SIGCOMM "Test of Time" award for the paper "PlanetLab: An Overlay Testbed for Broad-Coverage Services"

Gustavo gave the talk "The Case for Customized System Development" as one of the keynotes in the 7th ACM International Systems and Storage Conference (Systor) in Haifa, Israel.

Abstract:

The IT industry is undergoing a number of substantial changes that go beyond the "technology revolutions" that marketing strategies periodically generate. The shift in computer systems' scale, the move towards a service model, the total cost of ownership for IT systems, and the added stringent requirements on these systems have changed the predominant balance of the last decades. Instead of general purpose solutions, the trend in system development these days is toward new degrees of customization of both hardware and software. Such customization is predominantly guided by hardware developments, which are introducing significant constraints on how programs run but also opening many new possibilities for those who can exploit them. In this talk, I discuss the trend towards custom design, explain why it is happening, present several examples, and give an overview of the opportunities and challenges it creates for systems research.

This summer again several PhD students will be widening their knowledge and experience through their internships. Below is the list of students and companies they are visiting:

Jana Giceva - Microsoft Research, Silicon Valley

Simon Gerber – Hewlett Packard, Silicon Valley

Zsolt Istvan - Microsoft Research, Redmond

Lucas Braun – Microsoft Research, Redmond

Darko Makreshanski – Microsoft Research, Redmond

Gerd Zellweger –Microsoft Research, Redmond (starting mid June)

Stefan Müller - Amazon (starting July)

Tahmineh Sanamrad is one of the prestigious Google Anita Borg Scholarship Award recipients. The award supports women in computing and technology.  More about Anita Borg Scholarship Award here.

The paper entitled "Grok the Data Center" by Zaheer Chothia, Qin Yin and Timothy Roscoe has been accepted to the 5th ACM SIGOPS Asia-Pacific Workshop on Systems (APSys 2014), June 25-26th, 2014 in Beijing, China.

Martin Kaufmann defended his PhD Dissertation entitled: Storing and Processing Temporal Data in Main Memory Column Stores

Committee: Prof. Dr. Donald Kossmann, ETH; Prof. Dr. Gustavo Alonso, ETH; Prof. Dr. Christian Jensen, Aarhus University, Denmark; Dr. Norman May, SAP, Germany.

Louis Woods defended his PhD Dissertation entitled: FPGA-Enhanced Data Processing Systems

Committee: Prof. Dr. Gustavo Alonso, ETH;  Prof. Dr. Donald Kossmann, ETH; Dr. Ken Eguro, MSR; Prof. Dr. Jens Teubner, TU Dortmund.

Louis will join Oracle Research, Redwood Shores (USA) this summer. 

Spyros Voulgaris has gotten Tenure at the Computer Science Department of VU Amsterdam, The Netherlands. Spyros was a post-doc in the Systems Group from 2006 to 2008 before he joined VU Amsterdam. Congratulations, Spyros!

Gustavo Alonso gives an overview of new trends and job profiles in IT profession in the Netzwoche article entitled: "Berufswandel in der IT: Düstere Zukunft für Quereinsteiger".

Cagri Balkesen defended his PhD Dissertation entitled:  In-Memory Parallel Join Processing on Multi-Core Processors

Committee: Prof. Dr. Gustavo Alonso, Prof. Dr. Donald Kossmann, Prof. Dr. Tamer Ozsu (U. of Waterloo),  Dr. Stefan Manegold (CWI, the Netherlands).

Cagri will join Oracle Labs (Zürich) in July.

The paper "A Hash Table for Line Rate Data Processing" by Zsolt Istvan, Gustavo Alonso, Michaela Blott (Xilinx), and Kees Vissers (Xilinx) has been accepted for publication by the journal ACM Transactions on Reconfigurable Technology and Systems (ACM TRETS).

The following paper has been accepted at SIGMOD 2014, Snowbird, Utah, USA, June 22-27, 2014:

“Towards Indexing Functions: Answering Scalar Product Queries”. Arijit Khan, Pouya Yanki, Bojana Dimcheva, and Donald Kossmann.

Abstract
“In this study, we consider a broad category of OLAP queries which can be expressed as the scalar product between a known expression (function) over multiple database attributes and an unknown set of parameters. Scalar product queries naturally arise in a wide range of applications including moving-objects intersection finding, time-series prediction, scientific simulation, and in active learning. We design a lightweight, yet scalable, dynamic, and generalized indexing scheme, called the Planar index, for answering such scalar product queries in an accurate manner.”

Timothy Roscoe is delivering the opening keynote talk entitled “Minding the Gap” at the Conférence d’informatique en Parallélisme, Architecture et Système (ComPAS) in Neuchâtel, Switzerland.

Abstract:
There is a (by now) well-established gap between the functionality provided by hardware architects, and the issues and requirements that system software (and particularly operating system) designers worry about. Recently, with the accelerating pace of hardware innovation, this has become a serious and pressing issue, and I will argue that the way we have traditionally built low-level system software is simply not up to the challenge of supporting modern, diverse, highly parallel, rapidly evolving, and quirky hardware.

The Barrelfish research operating system was conceived as a response to trends in hardware design, and continues to serve as a testbed for new ideas in OS construction which apply modern techniques from programming languages, knowledge representation, and networking to the problem of managing the machine. I will describe a selection of these in the talk, together with our experiences implementing them. Our ideas have proven useful and effective at dealing with the mismatch between hardware architecture and software needs, but ultimately we hope they will foster a more fruitful conversation between the hardware and software sides of "the gap", and lead to better principles and practices on both sides.

A paper describing Pydron, Stefan Müller's parallelizing tool for Python code, has been accepted for publication at IEEE Computer in a special issue about computing in astronomy.

Scaling Astroinformatics with Pydron

Astronomers like working directly with their data using scripting languages such as Python. However, huge data volumes now force scientists to run their code on clusters and clouds. Pydron helps to bridge the gap between interactive data analysis and scalable computing resources.

The paper "Operating system support for planetary-scale network services" co-authored by Timothy Roscoe received the first NSDI 10-year "Test of Time" award. The paper was published in 2004 at the first USENIX Symposium on Networked Systems Design and Implementation (NSDI 2004) held in San Francisco.

This is the second "Test of Time" award for the work on the same system. Last year Timothy Roscoe was one of the recipients of the 2013 ACM SIGCOMM "Test of Time" award for the paper "PlanetLab: An Overlay Testbed for Broad-Coverage Services".

David Sidler joined the Systems Group as a new PhD student. David completed his Master Thesis in our group.

On March 25th there was a mini-workshop covering the ongoing
collaborations between the Systems Group and Oracle Research together
with Thomas Würthinger, of the Oracle Labs in Linz, Austria.

Alessandra Loro joins the Systems Group as a new Post Diplomand. Alessandra completed her Master Thesis in our group in February.

The poster entitled "Rethinking the Unit of Scheduling of Parallel DB Operators" by Jana Giceva, Gustavo Alonso and Timothy Roscoe has been accepted at EuroSys 2014, which will be held in Amsterdam, April 13-16, 2014.

Abstract:

Scheduling of parallelized DB operators is both challenging and important problem for modern databases. The penalties for resource sharing are aggravated with the new operator algorithms tuned for efficiently exploiting the new hardware features. Hence, they must be considered more seriously now and in the future. By rethinking the unit of scheduling from CPU cores to, for instance NUMA regions, we can achieve less performance degradation due to sharing, and more efficient leverage of machine parallelism that will in turn maximize the overall workload throughput. Even though we are still at an early stage of evaluation, the initial results are encouraging and we believe that the approach is applicable to a wider range of database operators as well as other applications that have similar properties and requirements.

Pratanu Roy's paper entitled "Low Latency Handshake Join" is accepted at VLDB 2014 in Hangzhou, China, September 1-5, 2014.

Prataus work is based on the recently proposed handshake join algorithm, which is a mechanism to parallelize the processing of stream joins in a NUMA-aware and hardware-friendly manner. Handshake join achieves high throughput and scalability, but it suffers from a high latency penalty and a non-deterministic ordering of the tuples in the physical result stream... The paper will be published after the VLDB 2014.

Timothy Roscoe visted Google and HP Labs in Silicon Valley and gave the following talk at both companies:  Towards an intra-machine network architecture.

Selected as one of the best paper of ICDE 2013, Cagri Balkesen's work on main memory hash joins has just been accepted to the IEEE journal Transactions on Knowledge and Data Engineering.

Renato Marroquin joined the Systems group as a new PhD student. Renato completed his Master Thesis at Pontifical Catholic University of Rio de Janeiro, Brazil.

Gustavo Alonso gives a talk on "Heterogeneous Architectures in Data Centers" at the Xilinx Emerging Technologies Symposium in San Jose, California.

Ercan Ucan defended his PhD Dissertation entitled: Data Storage, Transfers and Communication in Personal Clouds.

Committee: Prof. Dr. Timothy Roscoe, ETH; Prof. Dr. Gustavo Alonso, ETH; Dr. Anne-Marie Kermarrec, Senior researcher at INRIA and "part time invited Prof." at EPFL

Marica Stojanov will join Credit Suisse on February 3rd, 2014.

Tudor Salomie defended his PhD Dissertation entitled: Cloud-ready scalable and elastic data processing using off-the-shelf databases, replication and virtualization.

Committee: Prof. Dr. Gustavo Alonso, Prof Dr. Timothy Roscoe, Dr. Tim Harris (Oracle, Cambridge, UK)

The following paper has been accepted at EDBT 2014, to be held March 24-28, 2014 in Athens, Greece.

Title: Fast Reliability Search in Uncertain Graphs

Authors: Arijit Khan (ETH), Francesco Bonchi (Yahoo!), Aristides Gionis (Aalto University), and Francesco Gullo (Yahoo!)

Abstract: Given an uncertain graph with probabilities on edges, we propose efficient indexing methods to find all nodes reachable from a query node with probability no less than an input probability threshold. Such reachability queries can be useful in influence maximization in social networks, predicting co-complex memberships in protein-interaction networks, and computing packet delivery probability between a pair of nodes in mobile ad-hoc networks.

The paper "Shared Workload Optimization" by Georgios Giannikis, Darko Makreshanski, Gustavo Alonso, and Donald Kossmann has been accepted for presentation at VLDB 2014. The paper describes a query optimizer for the SharedDB database engine developed in the group.

Eric Sedlar from Oracle Labs visited the System Group. A small workshop took place on 3rd December 2013 to discuss ongoing research around data processing on modern hardware.

Tudor Salomie presented his work on virtualized and replicated
databases at the IMDEA Networks Institute in Madrid, Spain

Thomas Etter took up the positon of a Software Ingineer at Google, Zürich. Thomas will start at Google on Dec. 2. 

Kornilios Kourtis presented the following paper at PLOS 2013 (7th Workshop on Programming Languages and Operating Systems), Nemacolin Woodlands Resort, Pennsylvania, USA, Nov. 3, 2013: 

"Modeling NICs with Unicorn" by Pravin Shinde,  Antoine Kaufmann, Kornilios Kourtis and  Timothy Roscoe.

The following paper will be published in ACM Transactions on Database Systems:

"XLynx—An FPGA-based XML Filter for Hybrid XQuery Processing" by Jens Teubner, Louis Woods, and Chongling Nie.

Claude Barthels joins the Systems Group tomorrow (Oct. 1st) as a new PhD student. Claude completed his Master Thesis in our group.

Arijit Khan joined the Systems Group today (9.11.2013) as a postdoctoral researcher. Arijit is coming to our Group from University of California, Santa Barbara, CA, USA.

The Systems Group is presenting the following papers and demos at VLDB and co-located workshops in Riva del Garda, Italy, August 26-30, 2013.

• Tahmineh Sanamrad. Donald Kossmann. Query Log Attack on Encrypted Databases. VLDB 2013. Secure Data Management Workshop, Riva del Garda, Italy, August 2013.
• Martin Kaufmann. Storing and Processing Temporal Data in a Main Memory Column Store. VLDB 2013 PhD Workshop, Riva del Garda, Italy, August 2013.
• Martin Kaufmann, Panagiotis Vagenas, Peter M. Fischer (Univ. of Freiburg), Donald Kossmann, Franz Färber (SAP). DEMO: Comprehensive and Interactive Temporal Query Processing with SAP HANA. VLDB 2013, Riva del Garda, Italy, August 2013.
• Martin Kaufmann, Peter M. Fischer, Norman May, Andreas Tonder, Donald Kossmann. TPC-BiH: A Benchmark for Bi-Temporal Databases. VLDB 2013. TPCTC 2013, Riva del Garda, Italy, August 2013.
• Anja Gruenheid, Donald Kossmann. Cost and Quality Trade­-Offs in Crowdsourcing. VLDB 2013. DBCrowd, Riva del Garda, Italy, August 2013.

The paper "Modeling the execution semantics of stream processing engines with SECRET" by Nihal Dindar, Nesime Tatbul, Renée J. Miller, Laura M. Haas, Irina Botan is published in VLDB Journal. Volume 22. Number 4. August 2013.

Marica Stojanov joins the Systems Group as a new PhD Student. Marica comes to us from EPFL Lausanne, where she completed her Master Thesis.

Pravin Shinde started his 3-month internship in the Systems research group at HP Labs in Paolo Alto, CA, USA. Pravin will be working on exploratory project involving figuring out how smart NICs can be used for the benefits of the applications.

The following demo was accepted to FPL 2013 (23rd International Conference on Field Programmable Logic and Applications), Porto, Portugal, 2-4 Septemeber, 2013:

Hybrid FPGA-accelerated SQL Query Processing by Louis Woods, Zsolt Istvàn, and Gustavo Alonso.

The following paper was accepted at VLDB 2014, Hangzhou, China, September 1-5, 2014.

Multi-Core, Main-Memory Joins: Sort vs. Hash Revisited by C. Balkesen, G. Alonso, J. Teubner (TU Dortmund University), T. Özsu (University of Waterloo).

Gustavo Alonso has been recognized for his research in distributed systems, middleware, and data management at the ACM Fellow Ceremony in San Francisco, CA. Interview with Gustavo Alonso 

Markus Gross, Vinton Cerf and Gustavo Alonso at the ACM annual Awards Banquet

The book "Data Processing on FPGAs" published by Jens Teubner and Louis Woods gives an introduction to field-programmable gate arrays targeted at a database audience. Besides explaining core FPGA technology and programming models, the book also discusses several database use cases that benefit from FPGAs, including stream processing applications, core database operators, and security sensitive applications.

An interview with Donald Kossmann entitled "Big Data, big risks?" was published in ETH Life. The interview is available here (in German).

The Systems Group is presenting the following papers and demos at SIGMOD in New York, USA, June 22-27, 2013.

• Martin Kaufmann, Amin A. Manjili, Peter M. Fischer (Univ. of Freiburg), Donald Kossmann, Franz Färber (SAP), Norman May (SAP): Timeline Index: A Unified Data Structure for Processing Queries on Temporal Data.
• Giorgos Giannikis, Darko Makreshanski, Gustavo Alonso, Donald Kossmann. DEMO: Workload Optimization using SharedDB.
• Louis Woods, Jens Teubner, Gustavo Alonso. DEMO: Less Watts, More Performance: An Intelligent Storage Engine for Data Appliances.
   

Karolina Alexiou joins the Systems Group as a new Post Diplomand. Karolina completed her Master Thesis in our Group this spring.

A paper describing an implementation of Memcached on FPGAs, co-authored by Zsolt Istvan and Jeremia Baer, has been accepted at HotCloud 2013. The work is part of the ECC collaboration between Xilinx and the Systems Group.
 

Ercan Ucan presented the following paper at NETYS in Marrakesh, Marocco, May 2-4, 2013.

Establishing Efficient Routes between Personal Clouds; Ercan Ucan, Timothy Roscoe.

The paper "Parallel Computation of Skyline Queries" by Louis Woods, Jens Teubner, and Gustavo Alonso got the best paper award at FCCM'13, Seattle, USA.

Zsolt István joined the Systems Group today as a new PhD student. Zsolt completed his Master Theisis in our group in March, 2013.

Nesime Tatbul has joined Intel Labs as a senior research scientist to work as part of the Intel Science and Technology Center for Big Data based at MIT CSAIL.

Gustavo Alonso gives a keynote talk, entitled “Hardware Killed the Software Star”, at ICDE 2013, in Brisbane Australia.

The "Time Travel in Column Stores" paper by Martin Kaufmann, Amin Amiri Manjili, Stefan Hildenbrand, Donald Kossmann, Andreas Tonder (SAP) was presented by Martin Kaufmann at ICDE 2013 in Brisbane, Australia.

Martin also presented the following demo: "A Generic Database Benchmarking Service" by Martin Kaufmann, Peter M. Fischer (Univ. of Freiburg), Donald Kossmann, and Norman May (SAP).

Gustavo Alonso has presented several of the projects around SwissBox at VU Amsterdam and at IBM Rüschlikon.

Timothy Roscoe will be in sabbatical at University of Washington through September 2013.

Cagri Balkesen has given a talk at Oracle Research presenting his work on implementing main memory relational joins on multicore architectures.

Alexandru Moga took up the position of R&D Computer Scientist with ABB Corporate Research in Baden-Daettwil. He will start at ABB in April 2013.