Proposals

Linux Kernel Scalability for Systems with Large Number of CPUs

This proposal has been rejected.

*

One Line Summary

We will discuss the issues and lessons we have learned scaling Linux to systems with 64 or more cpu cores, and work that still remain.

Abstract

As the number of Cores for the current generation of CPU reach double digit, systems that have number of CPU cores in the hundreds will become common in the near future. Locking bottlenecks could easily occur on large system if the code design did not have scalability on such scale in mind. We will discuss scalability issues we have found in file systems, I/O, memory management to provide lessons on practices to avoid, and also areas in the kernel where work still need to be done.

Tags

performance, scalability

Speakers

  • Tim Chen

    Intel

    Biography

    Tim Chen currently work at Intel’s Open Source Technology Center. He works on the kernel performance project to tune and improve the performance of Linux kernel. He was a graduate from UCLA.

  • Andi Kleen

    Intel

    Biography

    Andi Kleen has been contributing to the Linux kernel since the mid 90ies.
    He worked on many different areas, including network stacks, VM, scalability,
    NUMA, performance tuning, file systems, error handling and low level
    architecture code.

    He was the initial maintainer of the x86-64 Linux port, nursing it from
    obscurity to wide deployment and also mainted the 32bit x86 port for
    a few years.

    Andi spent nearly 10 years in SuSE Labs at SUSE, later Novell as a kernel
    engineer. He joined the Intel Open Source Technology Center in 2008 where
    he is working on reliability and scalability for Intel servers.

    His other software interests include
    compilers and cryptography. He also enjoys learning about biology.

Leave a private comment to organizers about this proposal