Comments on: MySQL performance on Amazon EC2

By: Thorsten

Thorsten — Tue, 01 Dec 2009 08:01:29 +0000

Greg, I haven’t seen “hard data” on this. There are variations, which are pretty tight nor normal use, but for benchmarking it can be challenging.

By: Greg

Greg — Tue, 10 Nov 2009 01:05:22 +0000

We’re looking to set up automated performance testing of our service and wondered whether we could run such benchmarks on EC2 instead of our own controlled boxes. I’m really curious, is there any hard data either way on whether instances provide consistent performance?

Thanks

By: Blog Joyent | On Benchmarking Databases: MySQL on Joyent versus AWS (part 1)

Mon, 17 Aug 2009 12:48:37 +0000

[...] OTLP benchmark and with both it’s read-write and read-only tests. There is some existing, but rudimentary benchmark numbers using Sysbench’s OTLP from Rightscale. So we could at least know that we were on the right [...]

By: Guy Rosen

Guy Rosen — Tue, 28 Jul 2009 14:03:26 +0000

The community here might like to see some of our own results benchmarking MySQL on EC2 – including a fascinating finding on how CPU types affect the performance you get!
http://www.infibase.com/blog/2009/07/mysql-on-amazon-ec2-part-1/

By: Understanding Cloud Benchmarks | Union Station

Understanding Cloud Benchmarks | Union Station — Tue, 26 May 2009 15:01:05 +0000

[...] is at least within the same ballpark as most commodity [...]

By: Thorsten

Thorsten — Sat, 04 Apr 2009 13:42:07 +0000

Nick, this is very interesting. Do you have any data to back up your claim, or is it just a rant? I’m sure you’ve done some benchmarking, perhaps you care to share?

By: Nick

Nick — Tue, 31 Mar 2009 20:41:00 +0000

I have found that the CPU performance is irregular at best.

My finding is that running a CPU based benchmark on EC2 instances varies, and you will not receive consistent CPU performance over time. The variations in performance were very evident across all instance types.

If you are CPU bound, make sure you understand how irregular the performance of EC2 really is.

By: Thorsten

Thorsten — Tue, 17 Mar 2009 06:00:52 +0000

Dave, I don’t have benchmarks I can publish for this. My rules of thumb are: if you need bandwidth use the local drives striped. If you need I/O ops (i.e. random access) use EBS but watch the cost. You can stripe across a couple of EBS volumes and I know of people who do that, but I’d work really hard at somehow splitting up the load to avoid it, it just seems like pushing the envelope too far.

By: Slicing the cloud at The Relenta blog

Slicing the cloud at The Relenta blog — Fri, 13 Mar 2009 18:57:31 +0000

[...] Amazon EC2 (as reported by RightScale): [...]

By: Dave

Dave — Fri, 13 Mar 2009 17:03:18 +0000

RE: bottlenecks – I’d also be interested to see the results of I/O bound benchmarks. We’re in the process of migrating a dedicated server that’s currently I/O bound to EC2, would be nice to see someone else’s conclusions.