Here are the slides to my first talk at Percona Live UK 2011: Choosing a MySQL High Availability solution.1
- 1. See this for a review of the conference as a whole: https://openlife.cc/blogs/2011/october/thanks-percona-and-attendees-gre…]
Choosing a MySQL High Availability solution - Percona Live UK 2011
The main point of feedback I got from this talk was that people appreciated that I had said they shouldn't deploy clustering frameworks that they don't understand. Precious advice indeed!
One member of the audience also came to discuss that they can't use Galera if it reduces the maximum performance by 50% (referring to the disk bound workload). I asked what HA they currently used, and it was MySQL replication. I then commented that they don't get any better performance with that solution either, since MySQL replication is a huge bottleneck, and that for all I know Galera has better throughput than that in all cases. This is good to remember when reading these benchmarks: I'm trying to find out the overhead of different replication solutions by comparing them to the performance of a single server without replication. But this doesn't mean that whatever you run in production would actually have that maximum performance anyway.
Even so, that comment and the recent benchmarks by Tungsten boasting good performance also on a disk bound workload have tickled my curiosity enough. I think I'll have to return to this disk bound Galera test once more to see if I can tune InnoDB to give me better performance. AFAICS there is no reason in Galera itself why it wouldn't be achievable.
The slides are also attached here in both PDF and ODP format.