Scaling MySQL on a 256-way T5440 server using Solaris ZFS and Java 1.7
(and then wipe coffee off the computer)
of course the real aim should be to scale with one instance on the machine as scaling with multiple instances on the one machine isn’t scaling at all – it’s scale out, but with more problems (now when one machine goes down, so do 1110202434 database instances).
Hey Stewart, I spoke to Benoit at the CEC last week. He’s interested in testing the ndbmtd and/or several ndbd nodes together on that system as well.
yeah..my thoughts exactly. When I saw this initially I thought that is what they had done (scale one instance). Maybe someone in Sun is working on that.
First, I would like to say that there is single instance performance and scalability efforts going on at Sun. We have a large team working on this issue and progress is being made already. Allan Packer is leading the team that is working on this and his blog (http://blogs.sun.com/allanp/) sometimes has information on the performance project. One of the videos he references talk about the future of the performance project at Sun. (http://link.brightcove.com/services/player/bcpid1683878780?bctid=1892187731)
It should be noted that improving performance of an application is a long term effort. Performance gains come in small doses typically and this project is one that might take some time.
Given that, there are other people at Sun that are trying to figure out ways MySQL can be used on larger systems (like the T5440). The thought is around horizontally scalability within a system– either using virtualization or not. This article by Benoit was simply one example of these efforts. I understand the concerns with this approach (losing the system causes loss of multiple databases), but there advantages (like cost savings, energy savings, consolidaton, etc.) that play a part as well.
I’d be interested in hearing more of your thoughts on horizontally scaling within a system.
I’m sitting around in a room with a bunch of fellow expert MySQL hackers and the consensus is: what a bunch of bullshit.
Is there a tool that makes it easy to manage a cluster of MySQL servers running on a big SMP?
Well… you could use the Instance Manager but it was deleted in 6.0 because it was worked out that nobody ran multiple instances on the same box.
oh, there were one or two bugs as well :)