Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I'm wondering how to compare performance of EC2 instances with standard dedicated servers and desktop. I've found only comparance of defferent clouds.

I need to find a solution to perform some computations which require CPU and memory (disc IO is not used). The choice is to use:

  • EC2 (High-CPU) or
  • Xeon 5620/5630 with DDR3 or
  • Core i7-960/980 with DDR3

Can anybody help, how to compare their performance? I'm not speaking about reliability of alternatives, I want to understand pros and cons from the point of just performance.

share|improve this question

2 Answers 2

The sort of analysis you're looking for can be tricky to do. If you are worried purely on performance, and disk I/O is not a significant contributor to performance, then CPU and RAM performance will impact your overall performance much more. For that, you'll want to look at benchmarks that come close to the kinds of calculations you're looking to run. Different Intel processors perform percentage-points better than others for certain tasks and data-set sizes.

In general cloud computing performance runs very close to the actual CPU they're running on the back end. RAM performance can vary quite a bit based on what virtualization technology they're using and how loaded the individual CPU busses are with I/O across the entire node. EC2 RAM performance should be less consistent than physical hardware (but could be just as consistent, it varies).

Without knowing more details about the exact type of computation you're doing, that's as close to an answer as I can give you.

share|improve this answer

With EC2, you're not guaranteed performance. Due to the nature of virtual hardware, it all depends on the overall health of the cloud. If you're looking for something that delivers identical performance every time, most providers (Amazon included) offer what are called reserved instances.

Think of the on-demand instances are more of a, "If I need more, I can just launch more." I.e., if you're looking to process 100 events per second, but launching one only provides you with 75 events per second, you can simply launch another instance. Based on your question, it sounds like you need a dedicated server for continuous processing--or maybe even an offsite dedicated server (ala Rackspace).

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.