Assistant Professor
Tel: (+972) 77 887 1503
Office: Fishbach 408-5 (TCE)
Electrical Engineering Department
Technion
Haifa 32000
Israel
Research interests
My focus is on building computer systems with programmable computational accelerators: GPUs, FPGAs. I work on a broad range of systems projects on hardware architecture, programming tools, operating systems, security and privacy, high concurrency servers and smart I/O devices. All these projects strive to solve some of the challenges that complicate the design of efficient and fast systems with programmable accelerators. I am also interested in applications of accelerators to machine learning and high performance computing.
A high-level overview of my research agenda is available here: [video] [slides].
For a more detailed and somewhat forward-looking overview check out my HotOS paper [pdf] [slides]
In Spring 2018 I teach a new seminar together with Yoav Etsion and Shahar Kvatinsky
I also continue teaching my original course:
Check out my recent posts on
Selected publications
Google scholar profile- [USENIX ATC18] "Varys: Protecting SGX enclaves from practical side-channel attacks"
- O. Oleksenko, B. Trach, R. Krahn, A. Martin, M. Silberstein, C. Fetzer
- S. Kim, A. Wated, Y. Hu, X. Zhang, S. Huh, E. Witchel, M. Silberstein
Students
- MSc[2017- ] Marina Minkin, "Side channel mitigation for SGX enclaves"
- MSc [2017- ] Tanya Brokhman, "A unified page cache in heterogeneous systems"
- MSc [2016 - ] Lior Zeno, "Streaming applications on smart NICs"
- PhD [2016 - ] Meni Orenbach, "Efficient OS abstractions for GSX enclaves"
- PhD [2016 - ] Haggai Eran,"OS abstractions for network application accelerators"
- MSc [2016- ] Vasilis Dimitsas, "I/O prefetcher for GPUs"
- MSc [2016-2017] Shai Vaknin, "Distributed training of DNNs via direct GPU I/O"
- MSc [2015-2017] Shai Bergman "High performance Disk I/O on GPUs"
- MSc [2014-2017] Amir Wated, "High concurrency servers on GPUs"
- MSc [2014-2015] Mattan Hamilis (with Eli Ben-Sasson), "High Performance Additive FFT in finite fields" [now @ IDF]
- MSc [2014-2016] Feras Daud "High performance network I/O on GPUs" [ now @ Mellanox
- MSc [2014-2016] Sagi Shahar, "Support for mmap on GPUs" [ now @Google]
Research Projects
Ongoing
- High performance SGX enclaves
- OS abstractions for smart NICs
- Operating system services for GPU programs
- GPUnet: Network sockets for GPUs.
- GPUrdma: GPU-side Infiniband verbs
- GPUfs: File System support for GPUs. See my talk at GTC2012. Our ASPLOS13 paper is here. The source code is now available from GITHUB.
- Privacy in mobile systems
Production
EE046278/CS046278 - [ugrad/grad] Accelerators and Accelerated Systems
EE048991 Spring 2015 - [grad] Design and implementation of Deep Learning Systems
EE046209 Winter 2015,16 - [ugrad/grad] Operating systems structure
EE046274 Spring 2014,15,16 - [ugrad/grad] GPU-accelerated systems
EE048961 Winter 2014,16 - [grad] Selected topics in Operating Systems Design and Implementation
EE046274 Spring 2011 - [grad] Programming Massively Parallel Processors
CS236370 Autumn 2010 - [ugrad] Concurrent and Distributed Programming
CS236804 Spring 2010 - [grad] Seminar on parallel algorithms and applications for GPUs
CS236371 Spring 2010 - [ugrad] Project in parallel and distributed programming
Program committees
2018: ASPLOS
2017: SOSP, Middleware, MaRS (PC Chair and organizer)
2016: Systor (PC Chair), MaRS, PACT, VEE
2015: Systor, Eurosys
2014: Systor, ASPLOS external, SOSP poster
2013: IPDPS, CCGrid, SFMA, Systor
2012: CCGrid
2011: CCGrid
Short bio
I received my PhD in Computer Science at the Technion in 2010, under the supervision of Assaf Schuster and Dan Geiger.. My
thesis is on the mechanisms for building supercomputing systems using
large-scale distributed, unreliable resources and graphical processing
units (GPUs). I spent two years at the UT Austin as a post-doc with Prof. Emmett Witchel.
Family
Natalia Silberstein homepage
Boris Silberstein homepage
|
|
