The mission of the LTC Linux Performance
Team is making Linux better by improving Linux kernel performance, with
special emphasis on SMP scalability.
We measure, analyze, and improve the performance
and scalability of the Linux kernel, focusing on platform-independent issues.
To accomplish the mission and to achieve our goals, we use benchmarks selected
to provide coverage for workloads that align with IBM's Linux strategy
and with the objectives of the Linux Technology and Solutions Center (LT&SC):
scalable data center, carrier space and web server workloads. In addition,
we provide coverage for testing and exposing performance issues related
to specific Linux kernel components, such as the disk I/O subsystem.
Make Linux kernel run faster
Identify bottlenecks and improve scalability
in the Linux Kernel
Develop patches for improving performance/scalability
Provide a rational and scientific approach
to performance analysis
A microbenchmark for measuring the performance and scalability of the
I/O. A raw and direct I/O microbenchmark
A file system benchmark for measuring file I/O performance.
A microbenchmark for measuring system APIs latency and bandwidth
A file server benchmark for measuring how well a server can handle file
I/O requests from different client types.
A network benchmark for measuring network throughput on different
A client/server benchmark for measuring the performance of Java Enterprise
Application Servers using a subset of J2EE.
SDET. A development benchmark for measuring how a system handles a
development environment with a large number of users.
A benchmark for measuring the performance of NFS.
A benchmark for measuring the performance of World Wide Web Servers.
A file system benchmark for measuring multithreaded I/O performance.
A benchmark for measuring the performance of Java chat rooms.
Mary Edie Meredith and D. Vianney, Jan 2004, Linux
2.6 Performance in the Corporate Data Center, LinuxWorld Conference &
Expo 2004, New York.
Sandra Johnson, B. Hartner, and B. Brantley, Jan
Linux kernel performance and scalability, http://www-106.ibm.com/developerworks/linux/library/l-kperf/
Duc Vianney, Hyperthreading
speeds Linux, Jan 2003, http://www-106.ibm.com/developerworks/linux/library/l-htl/
Duc Vianney, "Linux
kernels 2.2, 2.4, and 2.5 performance comparison," Birds of a feather
session, LinuxWorld Conference & Expo 2002, SanFrancisco.
Duc Vianney, Sandra Johnson, and Bill
Hartner, "Linux kernel performance
measurement and evaluation," Linux Technology Center Chat, LinuxWorld
Conference & Expo 2002, SanFrancisco.
Bill Brantley, and Bill Hartner, "Strategy for Improving Linux Kernel Performance
and Scalability", IBM Academy Conference on Performance Engineering Best
Practices, July, 2002.
Mala Anand and Bill Hartner, TCPIP
Network Stack Performance in Linux Kernel 2.4. and 2.5. [OLS
Peter Wai Yee Wong, et al., Improving
Linux block I/O for enterprise workloads. [OLS 2002 presentation,
format, postcript format].
Duc Vianney and James Phelan, Faster
apps on a better machine. Tune your IBM Developer
Kit for Linux, v1.3, for faster Java application performance,
May 2001, http://www-106.ibm.com/developerworks/ibm/library/i-tuning/
Ray Byant and Bill Hartner,
technology, threads, and scheduling in Linux. Patching
the kernel scheduler for better Java performance, Jan.
Ray Bryant, Bill
Hartner, et al., SMP Scalability Comparisons
of Linux® Kernels 2.2.14 and 2.3.99, August 2000, LinuxWorld 2000.
Ray Bryant and John Hawkes, Lockmeter,
informative spinlock instrumentation for the linux kernel, 8/28/2000.
Bill Clark - Manager
Venditti - Manager
Bill Buros - Tech Lead
Last updated: 20 February 2004