CODEWITHSUNDEEP
apachetomcatnginxjetty
agentgonzo
agentgonzo

Reputation: 3643

Apache/Tomcat/Jetty/Nginx performance on EC2

Can someone give me some guidance on whether these numbers are expected, low or good. I've got little experience in high-volume HTTP services so don't know if this is about the limit that I could expect or if I'm doing something wrong and it can be improved a lot.

I'm just running an apache2 vanilla install with serving a "hello world" example and a very simple servlet service running in both Tomcat/Jetty.

For all services we're getting 4,000-7,000 requests/s served. Is this expected or quite low and we've got something wrong in the configuration. They're all just numbers to me at the moment and I don't know whether they are good or not!

This is running on an EC2 xlarge instance, but I'd be interested to see how that compares with a standard "fairly powerful standalone server".

Thanks.

Upvotes: 1

Views: 10407

Answers (2)

peterbooth
peterbooth

Reputation: 11

@agentgonzo, As you possibly suspected, these numbers are lower than you can expect on an EC2 m1.large instance. The best source of up-to-date objective data that I'm aware of is the Tech Empower benchmarks at http://www.techempower.com/benchmarks

Currently they a show hello world JSON servlet on Resin doing 35,000 requests/sec on EC2 versus 220,000 requests / second on a Sandy Bridge server.

On a new Ivy Bridge Core i7-3770 I get 325,000 requests / sec.

So yes, your results appear to indicate that there's a problem with your configuration or test approach.

Upvotes: 1

jesse mcconnell
jesse mcconnell

Reputation: 7182

Give this a good read, it ought to get you going:

http://therichwebexperience.com/blog/greg_wilkins/2010/06/lies_damned_lies_and_benchmarks

also, the days of having to put apache in front to protect java servlets ended years ago, especially when using things like jetty continuations or the async servlet mechanisms of servlet 3.0

Upvotes: 3

Related Questions