EC2 Instance Performance

This week I was able to do some load testing against a bunch of the EC2 instance sizes. The great news is that the instance performance we’ve all gotten to know around here (150/250/350 for small/large/x-large) has greatly improved with Amazon’s current infrastructure and the new Wowza code. Here’s the nutshell version, full details and methodology are on my blog at http://blog.ianbeyer.com/2013/02/26/wowza-ec2-capacity-update/.

m1.small: 250 Mbps

m1.medium: 500 Mbps

c1.medium: 950Mbps (full gigabit!)

m1.large: 700 Mbps

m1.xlarge: 950 Mbps

m2.xlarge: 300 Mbps

c1.xlarge: 950 Mbps

m2.2xlarge: 950 Mbps

Test was run from softlayer in Washington DC to us-east-1, which is approximately 2ms and 14 hops away.

Disclaimer: these numbers are by no means authoritative or official, but they’re a rough guideline on what you can expect from those instance sizes. Your mileage may vary, void where prohibited.

Hi ,

thanks for the effort, that makes thing more clear.

I have a question for you concerning the load balancing. Supposing that I lunch 2 instances c1.xlarge and use the EC2 load Balance Should this work ?.

If I launch a wowza load Balance with the same 2 c1.xlarge. How the bandhwith charge will be taken by those machines ?.

Thanks for your help

Mahmoud

The only additional load that you’ll get from the edges is the connection between them, so for all practical purposes, the increase in aggregate bandwidth is linear with the number of edges. The load balancer itself has negligible impact.

Typically what I’ll do is my origin server will also have the load balancer and and edge application, and then I can add edges at will.

thanks for your reply