Blog icon

How to shave 150ms off content served from EC2

By Glenn Davy,
Glenn Davy
Scroll down to read

It was a bit like Christmas morning when I opened my AWS management console, and noticed "Asia Pacific (Sydney)" had appeared on the menu of regions. 2 or 3 weeks before the circulating rumours had suggested! Many of our clients are deployed on ec2, and this should give us the chance to save our Australian clients the 150-200ms tax they pay per request to the US. But will it?

We couldn't help our selves, and span up a t1.micro in the new ap-southeast-2 region, one on the US West Coast, and one in Singapore, just to see what we could expect. From the New England tablelands a few hundred km north west of Sydney, the ping to the Sydney instance is 30ms. California was 190ms and Singapore 220ms.

But, we're interested in the web more than icmp packets, so I dropped nginx on to each instance, and hit the default 358 bytes welcome page, which takes no more than 1ms to serve. The latency to AWS Sydney was 43ms (total request/response cycle 44ms), the latency to California 198ms (total request/response cycle 199ms) and Singapore was quite random, with latencies of between 206ms and 773ms.

So if you are in our Great Southern Land and using AWS, migrating to use the Sydney data centre is definitely a good move.

Obviously, if you have a Ruby on Rails application you would like to migrate to the AWS Sydney datacentre, get in touch

blog comments powered by Disqus