Select event terms to filter by
« Week of February 26, 2018 »
Start: 01.03.2018 16:00


COMPASS: Computing Platforms Seminar Series

CAB E 72

Speaker: Saughata Ghose, Carnegie Mellon University

Title: How Safe Is Your Storage? A Look at the Reliability and Vulnerability of Modern Solid-State Drives






We live in an increasingly data-driven world, where we process and store a much greater amount of data, and we need to reliably keep this data around for a very long time. Today, solid-state drives (SSDs) made of NAND flash memory have become a popular choice for storage, as SSDs offer high storage density and high performance at a low cost. To keep up with consumer demand, manufacturers have been using a number of techniques to increase the density of SSDs. Unfortunately, this density scaling introduces new types of errors that can seriously affect the reliability of the data, and in turn significantly reduce the lifetime of the SSD.

In this talk, I will cover several issues that we have found which affect data reliability and vulnerability on modern SSDs available on the market today. I will explore two such issues in depth, along with solutions we have developed to mitigate or eliminate these issues. First, I will discuss read disturb errors, where reading one piece of data from an SSD can introduce errors into unread pieces of data. Second, I will discuss program interference errors, where writing one piece of data to an SSD can introduce errors both into other pieces of data and to data that has yet to be written. Notably, our findings show that the predominant solution adopted by industry to mitigate program interference actually introduces other interference errors, and exposes security exploits that can be used by malicious applications. For both issues, I will discuss solutions that we have developed based on these error types, which can buy back much of the lost lifetime, and which can eliminate the security exploits.


Saugata Ghose is a Systems Scientist in the Department of Electrical and Computer Engineering at Carnegie Mellon University. He received dual B.S. degrees in computer science and in computer engineering from Binghamton University, State University of New York, and the M.S. and Ph.D. degrees from Cornell University, where he was the recipient of the NDSEG Fellowship and the ECE Director’s Ph.D. Teaching Assistant Award. He received the Best Paper Award from the DFRWS-EU conference in 2017, for his work on recovering data from solid-state drives. His current research interests include application- and system-aware memory and storage systems, virtual memory management, architectural solutions for large-scale systems, GPUs, and emerging memory technologies. For more information, see his website at


Start: 02.03.2018 10:00

Speaker: Brad Beckmann (AMD Research)

Title: Processor Design for Exascale Computing

Date and Venue: Friday 2nd of March, 2018, at 10:00am, CAB E 72


The US Department of Energy’s exascale computing initiative aims to build supercomputers to solve a wide range of HPC problems, including emerging data science and machine learning problems. The talk will first cover the requirements for exascale computing and highlight various challenges that need to be addressed. The talk will then give an overview of the various technologies that AMD is pursuing to design an Exascale Heterogeneous Processor (EHP), which will serve as the basic building block of an exascale supercomputer. Finally, the talk will conclude by highlighting some of the simulation infrastructure used to evaluate EHP and our effort to open source and share it with the broader research community. Short Bio:

Brad Beckmann has been a member of AMD Research since 2007 and works in Bellevue, WA. Brad completed his PhD degree in the Department of Computer Science at the University of Wisconsin-Madison in 2006 where his doctoral research focused on physical and logical solutions to wire delay in CMP caches. While at AMD Research, he has worked on numerous projects related to memory consistency models, cache coherence, graphics, and on-chip networks. Currently, his primary research focuses on GPU compute solutions and broadening the impact of future AMD Accelerated Processing Unit (APU) servers. Regards, Juan Gómez Luna

Start: 02.03.2018 12:15