Rack-scale data processing system

Brief project description

We are building a data appliance for Rack-scale Computers (RaSC) that leverages the benefits of cross-layer optimization and provides support for  heterogeneous workloads. To achieve that we separate the storage from the data processing layer. The two layers communicate over a scalable interconnect fabric, at the moment focusing on RDMA over InfiniBand.

           

Both layers are internally distributed and are composed of heterogeneous components. For example the goal is that the processing layer can have a DBMS that uses the storage layer, but could also run machine learning or graph processing algorithms. 

This is an ongoing project that touches upon topics such as:

  • Internals of DBMS: storage engine components, optimizer, compiler, transaction manager
  • Development of DB algorithms on modern hardware
  • Resource management and scheduling on modern hardware
  • Distributed systems
  • Efficient networking over low-latency high-bandwidth interconnects
  • etc.

Therefore, there are many possibilities for interesting topics for BSc and MSc thesis and semester projects. Please contact us for most recent list of project ideas.

 

Project Members:

Past Project Members:

  • Jana Giceva (PhD Student)
  • Darko Makreshanski (PhD Student)
  • Daniela Dorneanu (PhD Student)
  • Alessandro Dovis (Master Student)
  • Stephan Dollberg (Master Student)

 

Publications

Papers

  • SwissBox: An Architecture for Data Processin Appliances -- CIDR 2011 (pdf)
    Gustavo Alonso, Donald Kossmann, Timothy Roscoe
  • Rack-scale data processing system -- WRSC 2015 (pdf)
    Jana Giceva, Darko Makreshanski, Claude Barthels, Alessandro Dovis, Gustavo Alonso
  • Work-sharing data processing systems -- PhD dissertation 2014 (pdf)
    by Georgios Giannikis
  • Deployment of Query Plans on Multicores -- VLDB 2014 (pdf)
    Jana Giceva, Gustavo Alonso, Timothy Roscoe, Tim Harris
  • Shared Workload Optimization -- VLDB 2013 (pdf)
    Georgios Giannikis, Darko Makreshanski, Gustavo Alonso, Donald Kossmann
  • Workload optimization using SharedDB -- SIGMOD demo (pdf)
    Georgios Giannikis, Darko Makreshanski, Gustavo Alonso, Donald Kossmann
  • COD: Database/Operating System Co-Design -- CIDR 2013 (pdf)
    Jana Giceva, Tudor-Ioan Salomie, Adrian Schupbach, Gustavo Alonso, Timothy Roscoe
  • SharedDB: Killing One Thousand Queries With One Stone -- VLDB 2012 (pdf)
    Georgios Giannikis, Gustavo Alonso, Donald Kossmann
  • Database Engines on Multicores, Why Parallelize When You Can Distribute? -- EuroSys 2011 (pdf)
    Tudor-Ioan Salomie, Ionut Emanuel Subasu, Jana Giceva, Gustavo Alonso
  • Predictable Performance for Unpredictable Workloads -- VLDB 2009 (pdf)
    Phillipp Unterbrunner, Georgios Giannikis, Gustavo Alonso, Dietmar Fauser, Donald Kossmann

Posters & Presentations

  • Eurosys poster 2015 (pdf)
  • Workshop for Rack-scale Computing (WRSC) presentation 2015 (pdf)