Linux on Power Performance Home
|
Below we consolidate pointers to descriptive information related to Performance for Linux on Power systems.
In general, we recommend
- Use the latest distro versions, service levels and updates available for the best performance and system behavior.
- In usage, your partition should be enabled for sharing CPU cycles and setup to allow your partition to take advantage of unused CPU cycles from other partitions
Contents
|
|
General information
Power.org
For a light read, check out the latest Power Instruction Set Architecture document 
. Before you click on this, the PDF file is over 1300 pages long..
- Around Page 272, there's Chapter 7. Vector-Scalar Floating-Point Operations (VSX) support
Understand your application
Before diving in with random performance improvements and tuning tips, it's important that you first understand the application, how it was designed, and where the bottle-necks may be.
We are in the process of developing content for the April 2009 Common conference in Reno which introduces new system admins to what's available on Linux on Power. This should be posted here in early May.
http://www.ibm.com/developerworks/wikis/display/LinuxP/Understanding+Linux+on+Power+performance
Related Performance Products
- MicroQuil's SmartHeap product
Performance monitoring
Materials available for performance monitoring include
- Live monitoring of memory, CPU, and I/O usage (vmstat, top, mpstat, iostat, nmon, etc)
- Tools for recording performance statistics (sar, nmon)
- Advanced tools for processor utilization resource register reports
Performance tools - problem determination
Performance problem determination
Performance insights
Performance insights
Performance tuning
Performance tuning
IBM High Performance Computing (HPC and Clusters)
HPC Central - Red Hat
HPC Central - Red Hat
HPC Central - Novell SUSE
New! IBM HPC Open Software for SLES 10 sp2
HPC Central - SUSE
Emerging Projects
