logo epfl
Ecole Polytechnique Fédérale de Lausanne
français | english
 EPFL > people@EPFL > Nikola Knezevic login

Doctoral Assistant
Distributed Programming Laboratory
Nikola Knezevic
web site: http://ch.linkedin.com/in/nikkne

office(s): INR312
phone(s): [+41 21 69] 38123
RESEARCH
My current research focuses on abortable Byzantine Fault Tolerant (BFT) protocols, their analysis and performance optimizations.

My main interests are: system programming, distributed and parallel computing, machine learning, and security.
BIOGRAPHY
Nikola Knežević received his Bachelors degree in Computer Engineering and Informatics from ETF, Belgrade in 2006. In the period of 2001-2003 he was working on maintaining Perl for Windows 95 platform. Currently, he is working towards obtaining his Ph.D. in the School of Computer Sciences and Communications, EPFL, under the supervision of Prof. Rachid Guerraoui, at Distributed Programming Laboratory.
MISSION
My mission is to build robust, scalable distributed platforms, that allow end users to explore their ideas without thinking about the complexities of the systems they use. Moreover, I believe in utilizing machine learning techniques, given the scale of current systems, as well as the behaviour of a large number of users accessing any system, to build customized, pleasant experience for each user. Lastly, as my previous work have demonstrated, I seek to mitigate the cost of errors in deployment of large scale systems, and provide developers the tools to discover errors with the minimal effort.
PAST PROJECTS
At Networked Systems Laboratory (NSL), I have been working on the MX, a large scale internet emulator, with self-tuning capabilities. Our goal was to build an emulator which will allows users to test both end-host and in-network routing protocols, without requiring abundant amount of machines. Further, emulator had to be self-tuning; it should adapt to ongoing experiments, reaching maximal utilization of testbed machines.

Another project I was working on deals with the issues of performance tuning and debugging using machine learning techniques. ML approach is appealing since applications are ubiquitous and produce large amount of data, which is useful for automated reasoning. My goal was to reduce the time developer spends on discovering performance bugs and their causes.

Previously I was working on large-scale data dissemination, node clustering, process recording in kernel. Also, during my internship at NSL in 2006, I have made MirageFS - host based virtual file system for UML.
MAIN PUBLICATIONS

Knezevic, N., and Guerraoui, R. A High-Throughput Byzantine Fault-Tolerant Protocol. PhD thesis, Lausanne, 2011. [ DOI | Details | Link ]

Guerraoui, R., Knezevic, N., Quema, V., and Vukolic, M. Stretching BFT. Tech. rep., 2010. [ Details | Full Text ]

Yabandeh, M., Knezevic, N., Kostic, D., and Kuncak, V. Predicting and Preventing Inconsistencies in Deployed Distributed Systems. ACM Transactions on Computer Systems (TOCS) (2010). [ DOI | Details ]

Guerraoui, R., Knezevic, N., Quema, V., and Vukolic, M. The Next 700 BFT Protocols. In Proceedings of the 5th ACM European conference on Computer systems (2010). Best Paper Award. [ Details | Full Text | Link ]

Yabandeh, M., Knezevic, N., Kostic, D., and Kuncak, V. Predicting and Preventing Inconsistencies in Deployed Distributed Systems. ACM Transactions on Computer Systems 28, 1 (2010). [ DOI | Details ]

Knezevic, N., Schubert, S., and Kostic, D. Towards a Cost-Effective Networking Testbed. In Proceedings of the 4th Workshop on Real Overlays and Distributed Systems (ROADS) (Also appears in SIGOPS Operating Systems Review, Volume 43, Number 4, pages 66-71, December 2009) (2009). [ Details | Full Text | Link ]

Yabandeh, M., Knezevic, N., Kostic, D., and Kuncak, V. CrystalBall: Predicting and Preventing Inconsistencies in Deployed Distributed Systems. In 6th USENIX Symp. Networked Systems Design and Implementation (NSDI) (2009). [ Details | Full Text | Link ]

Yabandeh, M., Knezevic, N., Kostic, D., and Kuncak, V. CrystalBall: Predicting and Preventing Inconsistencies in Deployed Distributed Systems. In Proceedings of the 6th USENIX Symposium on Networked Systems Design and Implementation (NSDI '09) (2009). [ Details | Full Text | Link ]

Yabandeh, M., Knezevic, N., Kostic, D., and Kuncak, V. CrystalBall: Predicting and Preventing Inconsistencies in Deployed Distributed Systems. Tech. rep., 2008. [ Details | Full Text | Link ]

Guerraoui, R., Knezevic, N., Quéma, V., and Vukolic, M. The Next 700 BFT Protocols. Tech. rep., 2008. [ Details | Full Text | Link ]

Crameri, O., Knezevic, N., Kostic, D., Bianchini, R., and Zwaenepoel, W. Staged Deployment in Mirage, an Integrated Software Upgrade Testing and Distribution System. In Proceedings of the 21st ACM Symposium on Operating Systems Principles (SOSP) (Stevenson, WA, U.S.A., 2007). [ Details | Full Text | Link ]

Vratonjic, N., Gupta, P., Knezevic, N., Kostic, D., and Rowstron, A. Enabling DVD-like Features in P2P Video-on-demand Systems. In Proceedings of the SIGCOMM Peer-to-Peer Streaming and IP-TV Workshop (2007). [ Details | Full Text | Link ]

Skills
Perl, C/C++/C#, R, Java, Scheme, Haskell, Ruby, Mace...

FreeBSD and Linux kernel, distributed version control systems (DCVS).

BSD, Linux, Mac OS X, and (limited) Windows administration.


©2004-2012 Nikola Knezevic - EPFL, 1015 Lausanne - last updated : 2011-11-27 15:53:07
The owner of this page is fully responsible for its contents