About Dr Baptiste Lepers

During morning peak hour, when many people are commuting to work in the same direction at the same time, traffic jams occur. Something similar is true of a computer: when many applications are trying to access the same data simultaneously, processing speed can become very slow. Dr Baptiste Lepers' research consists of analysing and modifying the behaviour of computer applications to avoid such electronic "traffic jams", resulting in faster and improved performance.

The pace of hardware innovation is fast, with performance progressing rapidly and modern computers now being incredibly powerful. But software applications are not always able to keep up with this pace, so it is often very difficult for software applications to use modern computer hardware efficiently. This has important cost and environmental implications, because hardware that is idle still uses a massive amount of electricity. So the challenge is to enable software to keep up, by designing new algorithms that are fast enough to use modern hardware efficiently. This is essentially what my research aims to do.

Specifically, I aim to track inefficiencies in applications' use of new computer hardware and fix them, automatically improving performance.

When a company upgrades its hardware, it usually hopes to get better performance - perhaps the ability to handle more customers or to serve them more efficiently. But unfortunately, because of the limitations of software, performance improvements brought by new hardware alone are usually slim.

So my work aims to benefit companies by enabling them to use their hardware efficiently, while also benefitting their customers by delivering a much faster and smoother user experience - because an application that uses hardware efficiently is much faster than one that doesn't.

The ultimate aim is to be able to automatically 'tune' applications so that they can adapt to new hardware themselves.

I joined the University of Sydney in 2018. It's a great place to do research, with a stimulating work environment in a nice location.

Selected publications

Traffic management: a holistic approach to memory placement on NUMA systems. M Dashti, A Fedorova, J Funston, F Gaud, R Lachaize, B Lepers, ...ACM SIGPLAN Notices 48 (4), 381-394
The Linux scheduler: a decade of wasted cores. JP Lozi, B Lepers, J Funston, F Gaud, V Quéma, A Fedorova Proceedings of the Eleventh European Conference on Computer Systems, 1-16.
Thread and memory placement on {NUMA} systems: Asymmetry matters. B Lepers, V Quéma, A Fedorova2015 {USENIX} Annual Technical Conference ({USENIX}{ATC} 15), 277-289.
MemProf: A Memory Proļ¬ler for {NUMA} Multicore Systems. R Lachaize, B Lepers, V QuémaPresented as part of the 2012 {USENIX} Annual Technical Conference ({USENIX}{ATC} 12), 53-64.
Large Pages May Be Harmful on {NUMA} Systems.F Gaud, B Lepers, J Decouchant, J Funston, A Fedorova, V Quéma, 2014 {USENIX} Annual Technical Conference ({USENIX}{ATC} 14), 231-242.
Challenges of memory management on modern NUMA systems. F Gaud, B Lepers, J Funston, M Dashti, A Fedorova, V Quéma, ...Communications of the ACM 58 (12), 59-66.
Everything you always wanted to know about multicore graph processing but were afraid to askJ Malicevic, B Lepers, W Zwaenepoel, 2017 {USENIX} Annual Technical Conference ({USENIX}{ATC} 17), 631-643.
Efficient workstealing for multicore event-driven systems. F Gaud, S Geneves, R Lachaize, B Lepers, F Mottet, G Muller, V Quéma2010 IEEE 30th International Conference on Distributed Computing Systems 516-525.