Wednesday Jan 26, 2005
Wednesday Jan 26, 2005
My name is Eric Saxe.
I've been with Sun for 5 years, the last 3.5 of which i've been doing development on the Solaris Kernel. Most of my development efforts are dedicated towards streamlining the kernel's scheduling and dispatching policies. Scheduling policy dictates when threads run (priorities), and dispatching deals with the "where" (on which CPUs) part.
This has become especially interesting work over the last few years as various non-uniformities have crept their way into the system's memory and processor architectures. Non-uniformity in the memory side of things has blessed us with NUMA, and has more recently emerged on the processor side though Simultaneous Multi-Threading (SMT), Chip Multi-Processor (CMP), and threaded multi-core processor architectures.
I'm looking forward to talking more about how these non-uniformites can impact application and system performance, as well as how the Solaris kernel adapts in a future post.
But for now, welcome!
I've been with Sun for 5 years, the last 3.5 of which i've been doing development on the Solaris Kernel. Most of my development efforts are dedicated towards streamlining the kernel's scheduling and dispatching policies. Scheduling policy dictates when threads run (priorities), and dispatching deals with the "where" (on which CPUs) part.
This has become especially interesting work over the last few years as various non-uniformities have crept their way into the system's memory and processor architectures. Non-uniformity in the memory side of things has blessed us with NUMA, and has more recently emerged on the processor side though Simultaneous Multi-Threading (SMT), Chip Multi-Processor (CMP), and threaded multi-core processor architectures.
I'm looking forward to talking more about how these non-uniformites can impact application and system performance, as well as how the Solaris kernel adapts in a future post.
But for now, welcome!
Posted by benr on January 26, 2005 at 11:06 AM PST #
A quick search with Google came up with this page about Memory Placement Optimization. I read the white paper before (it's dated 2003), and thought it was quite interesting and had some useful data. Would be nice to see some similar measurements on (say) a 4-way Opteron or UltraSPARC IIIi server.
Are there any similar pages and articles to look out for?
Posted by Chris Rijk on January 27, 2005 at 02:36 AM PST #
Posted by Eric Saxe on January 27, 2005 at 01:50 PM PST #
Posted by Ping-Wu Zhang on April 23, 2005 at 07:14 PM PDT #
Posted by Eric Saxe on April 25, 2005 at 10:00 PM PDT #