Log in

No account? Create an account

Best Laid Plans, Innodb Plugin

« previous entry | next entry »
Mar. 23rd, 2009 | 12:47 pm

We had this plan.

We were going to show off the latest Innodb plugin.

We imported into the tree... got a benchmark and then...

We found a race condition in our code that showed up once we added the new plugin to our code.

Is the plugin faster?

Sure seems to be... but make up your own mind.

You can play with it here: https://edge.launchpad.net/drizzle/trunk/2009.03.962 (which includes Solaris/Sparc support now)

Curious about the numbers? Below is a current run of sysbench from one of our 16core systems. Very soon now Jay will have the public regression system available.

What will it do? Anyone who wants to follow our efforts on performance will be able too.

Want to see if we regress? If we are getting faster?

You will be able to do that :)

Link | Leave a comment |

Comments {4}

Interesting graph...

from: mingenthron
date: Mar. 23rd, 2009 10:35 pm (UTC)

Looking forward to the details. Slope is in the right direction for most of that concurrency.... hope to see more so I can interpret throughput soon!

Reply | Thread

Brian "Krow" Aker

Re: Interesting graph...

from: krow
date: Mar. 23rd, 2009 10:42 pm (UTC)

If you want anymore details we can provide. Jay should be blogging soon on how we come to these and how to reproduce them. We really want people to be able to confirm this stuff on their own... and we want to get more feedback :)

Reply | Parent | Thread

Re: Interesting graph...

from: mingenthron
date: Mar. 23rd, 2009 11:03 pm (UTC)

Yep, already looked on Jay's blog for details... not there yet. I'll give him a bit of time before pestering him on details.

Reply | Parent | Thread

Re: Interesting graph...

from: jaypipes.myopenid.com
date: Mar. 23rd, 2009 11:08 pm (UTC)

The above is for a 4-way Xeon with 16 cores. I'll blog soon about the details.

Reply | Parent | Thread