Comments on: A brief update on NUMA and MySQL http://blog.jcole.us/2012/04/16/a-brief-update-on-numa-and-mysql/ Geek, electronics nerd, database nerd, father of three. Sun, 19 May 2013 02:57:21 +0000 hourly 1 http://wordpress.com/ By: MySQL如何避免使用swap[copy] | 香草的技术博客 http://blog.jcole.us/2012/04/16/a-brief-update-on-numa-and-mysql/#comment-2216 Mon, 08 Apr 2013 10:16:16 +0000 http://blog.jcole.us/?p=763#comment-2216 [...] A brief update on NUMA and MySQL http://blog.jcole.us/2012/04/16/a-brief-update-on-numa-and-mysql/ [...]

]]> By: mmap的MAP_POPULATE标志妙用 | 非业余研究 http://blog.jcole.us/2012/04/16/a-brief-update-on-numa-and-mysql/#comment-2028 Sat, 19 Jan 2013 13:43:17 +0000 http://blog.jcole.us/?p=763#comment-2028 [...] 前段时间在学习MySQL和内核的时候，碰巧遇到twitter的dba老大在折腾numa不平衡问题的问题而写了二篇numa相关的文章，非常有技术含量。参见 这里 [...]

]]> By: Marcos http://blog.jcole.us/2012/04/16/a-brief-update-on-numa-and-mysql/#comment-1807 Fri, 30 Nov 2012 15:50:16 +0000 http://blog.jcole.us/?p=763#comment-1807 Hello Jeremy! I presume you’re aware, but just in case: Vadim went ahead and implemented these in Percona Server :) http://www.percona.com/doc/percona-server/5.5/performance/innodb_numa_support.html

]]> By: Joost Horstman http://blog.jcole.us/2012/04/16/a-brief-update-on-numa-and-mysql/#comment-1729 Tue, 02 Oct 2012 11:48:03 +0000 http://blog.jcole.us/?p=763#comment-1729 Very good and clear explanation.

Thanks!

]]> By: Gavin Towey http://blog.jcole.us/2012/04/16/a-brief-update-on-numa-and-mysql/#comment-1278 Tue, 29 May 2012 17:44:39 +0000 http://blog.jcole.us/?p=763#comment-1278 Hi Jeremy,

I’m interested to know the behavior of the random stalls you were experiencing. Could you describe what you observe when they happen?

]]> By: John B. http://blog.jcole.us/2012/04/16/a-brief-update-on-numa-and-mysql/#comment-1192 Fri, 27 Apr 2012 04:28:48 +0000 http://blog.jcole.us/?p=763#comment-1192 Very cool. Nice work.

]]> By: Jeremy Cole http://blog.jcole.us/2012/04/16/a-brief-update-on-numa-and-mysql/#comment-1050 Tue, 17 Apr 2012 14:28:17 +0000 http://blog.jcole.us/?p=763#comment-1050 Basically, the hardware is NUMA regardless of what the BIOS is set to. When you “turn off” NUMA in the BIOS you are just making it impossible for the OS to control anything related to NUMA. We’d rather control it appropriately at the OS level and, in the future, take specific advantage of it.

]]> By: Ariel http://blog.jcole.us/2012/04/16/a-brief-update-on-numa-and-mysql/#comment-1042 Mon, 16 Apr 2012 23:43:34 +0000 http://blog.jcole.us/?p=763#comment-1042 Switching to SUMA in the bios causes the memory to be interleaved at a low level.

This is bascially the same as the configuration above, except that *everything* is interleaved, even things that would actually perform better if they were not interleaved.

It’s better to leave it as NUMA, and only interleave those things that need it.

]]> By: Brian http://blog.jcole.us/2012/04/16/a-brief-update-on-numa-and-mysql/#comment-1040 Mon, 16 Apr 2012 23:09:29 +0000 http://blog.jcole.us/?p=763#comment-1040 Why is this preferable to switching the hardware from NUMA to SUMA in the BIOS?

]]> By: The MySQL “swap insanity” problem and the effects of the NUMA architecture « Jeremy Cole http://blog.jcole.us/2012/04/16/a-brief-update-on-numa-and-mysql/#comment-1035 Mon, 16 Apr 2012 20:21:58 +0000 http://blog.jcole.us/?p=763#comment-1035 [...] Update: Also read A brief update on NUMA and MySQL. [...]

]]>