[Building Sakai] Rutgers is now using innodb-plugin in production

Hedrick Charles hedrick at rutgers.edu
Mon Dec 19 19:20:58 PST 2011


When planning for this I tried looking at benchmarks. Most showed that later versions of the plugin seemed to do similarly to 5.5. There were differences, but minor. When you look at release notes for the plugin and 5.5, the list of performance improvements is nearly the same.

I would be happy to move to 5.5 as soon as we know it's safe. I'm going to consider it for our 2.9 beta testing, although I'm wondering whether the Percona version is worth considering. Knowing that Longsite is using it makes it a possible alternative. I'm actually willing to pay Percona for maintenance for one server if what they give is worthwhile. My problem is that when we need support it is typically for an emergency, and I'm not sure whether that would be covered under maintenance.


On Dec 19, 2011, at 6:46:05 PM, Robert Cohen wrote:

> 
> Did you evaluate mysql 5.5?
> From what I've read its supposed to have performance improvements over 5.1
> even with the innodb plugin.
> 
> 
> 
> On 20/12/11 5:54 AM, "Charles Hedrick" <hedrick at rutgers.edu> wrote:
> 
>> We've had two episodes of the database server going to 100%. Because some
>> of my tests have shown that innodb-plugin works better under some kinds
>> of load, we're moving to it. We also had a DNS problem that might have
>> contributed to this. It's been fixed, so if our problems don't come back
>> we may not be sure what caused it.
>> 
>> We're using old versions of mysql 5.1, simply because we haven't wanted
>> to upgrade in the middle of the year, 5.1.44 with innodb plugin 1.0.6.
>> We'll probably be moving from our Solaris-based server to a server on
>> Linux running the newest 5.1, in late January. It's the final part of our
>> new Xenserver / iSCSI virtual architecture. After looking at various test
>> results I've decided to run the new system with hypertheading on. Innodb
>> plugin should deal with the multicore scaling problems sufficiently for
>> that to make sense.
>> 
>> _______________________________________________
>> sakai-dev mailing list
>> sakai-dev at collab.sakaiproject.org
>> http://collab.sakaiproject.org/mailman/listinfo/sakai-dev
>> 
>> TO UNSUBSCRIBE: send email to
>> sakai-dev-unsubscribe at collab.sakaiproject.org with a subject of
>> "unsubscribe"
> 
> 



More information about the sakai-dev mailing list