Maxime Lorrillere Email & Phone Number

Vancouver, Canada Area

Vancouver, Canada

Worked on virtual containers (VMs/containers) placements on multicore NUMA systems:• Designed a parallel memory migration technique to migrate containers (LXC) across NUMA nodes• Developed a scheduler to pack a maximum number of containers (LXC) on a machine while ensuringa performance target• Increased machine utilization by up to 4× while ensuring the performance goalWorked on cross-program memoization techniques for data analytics papielines (like Apache Spark) and databases:• Designed a LLVM pass (C++11) to automatically cache the results of deterministic functions

Paris, France

I taught the following courses at Pierre and Marie Curie University and Polytech Paris-UPMC:• Linux kernel programming (Master's degree)• Security (Master's degree)• Operating Systems (Bachelor's degree)• C programming (Bachelor's degree)

Paris, France

Worked on efficiently reusing unused memory in virtual machines:• Designed in-kernel metrics used to detect inactivity in virtual machines• Implemented a remote victim cache for the Linux kernel page cache so that an active VM can send its victim pages to an inactive VM• Evaluated on QEMU/KVM virtual machines controlled through libvirt with micro-benchmarks, databases (TPC-C and TPC-H on PostgreSQL) and scientific applications (BLAST)• Improved performance of I/O intensive applications by up to 45% when there is an inactive VM on the same machine• Improved performance up to 10x compared to state-of-the art techniques such as memory ballooning• This work has been published to SYSTOR'2015• Code available here: http://github.com/mlorrillere/pumaI also supervised student internship and student projects.

Paris, France

Studied the feasibility of a Linux kernel cooperative page cache:• Implemented a proof-of-concept of a remote page cache between virtual machines• Evaluation with micro-benchmarks showed large performance improvements for workloads with non-sequential access patterns

Paris, France

Worked on the performance evaluation of a fault tolerant token ring algorithm:• Developed a platform to test distributed algorithms.• Implemented a fault tolerant token-ring algorithm on top of this platform.• Analyzed its performance using fault injection and network latency emulation.

Lyon Area, France

• Designed a cross-platform 3D pre-processing tool for a finite element analysis simulator in structural mechanics• Implemented the pre-processing tool in C++ using the Qt framework and OpenGL

Doctor Of Philosophy (Phd), Computer Science

Master Of Science (M.Sc.), Distributed Systems

Bachelor Of Science (B.Sc.), Computer Science