How can I benchmark and measure cloud computing latency?
April 11, 2011 2:40 PM Subscribe
I am looking to do a project on the effects of latency on cloud computing, specifically latency impacting on Amazon EC2 virtual machines and I am trying to get my head around it. EC2 is an Infrastructure As A Service cloud product. Unfortunately, I do not have much experience with it.
I haven't found alot on the internet on this area; has anyone seen any information on this subject?
Could anyone make suggestions about appropriate ways to do latency benchmarking in this area.
What kind of things ought I be looking to include in any benchmark tests.
How exactly do others go about measuring latency for Amazon EC2. Are there any software tools out there which will help.
I'm a little confused as to how benchmarks of any sort would be different for EC2 than they would be for any other server or environment.
posted by toomuchpete at 3:09 PM on April 11, 2011
posted by toomuchpete at 3:09 PM on April 11, 2011
There are many different types of latency around EC2, depending on what you're trying to measure, so the question is really like saying "I want to do a project on temperature of computers, how do I measure that?"
For instance, there's 'EC2 to the average external website, speed of return, given certain file sizes' as reported in
http://www.scribd.com/doc/33118932/Amazon-EC2-Performance-Report-June
Or, there's 'latency of disk requests inside the EBS storage facility of EC2', as reported in
http://af-design.com/blog/2009/02/27/amazon-ec2-disk-performance/
and there's 'latency of internal network requests between two EC2 instances in the same availability zone' as discussed in
http://alan.blog-city.com/has_amazon_ec2_become_over_subscribed.htm
and in fact there's an inexhaustible, infinite supply of benchmarks and analyses on the net about ec2 and latency that is trivially accessible with google; don't mean to be insulting, but it sounds like you really need to just dig into the basic requirements of what you're looking for, understand what that means, and then go looking for a while.
posted by felix at 3:13 PM on April 11, 2011 [1 favorite]
For instance, there's 'EC2 to the average external website, speed of return, given certain file sizes' as reported in
http://www.scribd.com/doc/33118932/Amazon-EC2-Performance-Report-June
Or, there's 'latency of disk requests inside the EBS storage facility of EC2', as reported in
http://af-design.com/blog/2009/02/27/amazon-ec2-disk-performance/
and there's 'latency of internal network requests between two EC2 instances in the same availability zone' as discussed in
http://alan.blog-city.com/has_amazon_ec2_become_over_subscribed.htm
and in fact there's an inexhaustible, infinite supply of benchmarks and analyses on the net about ec2 and latency that is trivially accessible with google; don't mean to be insulting, but it sounds like you really need to just dig into the basic requirements of what you're looking for, understand what that means, and then go looking for a while.
posted by felix at 3:13 PM on April 11, 2011 [1 favorite]
Response by poster: Thanks for your advice - 'm not too knowlegable about this area and it shows! What are the best ways of measuring latency on a server? If I were to put applications up on the VM and compare and measure against the same applications running on a LAN would that not be an indicator of latency. Has benchmarking been done on Amazon EC2?
posted by conrad101 at 3:20 PM on April 11, 2011
posted by conrad101 at 3:20 PM on April 11, 2011
felix's key suggestion, and one which I echo, is: dig into the basic requirements of what you're looking for, understand what that means, and then go looking for a while.. If all you're interested in is the cumulative effect on end-user experience of the various latencies involved, then comparing a local VM to an Amazon VM would do it. If you want a break-down of what's actually CAUSING that end-user effect, then you need to do reading such as suggested by felix.
But first, you must figure out what you want to measure! ;)
posted by coriolisdave at 3:29 PM on April 11, 2011 [1 favorite]
But first, you must figure out what you want to measure! ;)
posted by coriolisdave at 3:29 PM on April 11, 2011 [1 favorite]
I think the reader is talking about network latency.
read this:
http://www.networkperformancedaily.com/2008/06/latency_and_jitter_1.html
posted by roboton666 at 4:47 PM on April 11, 2011
read this:
http://www.networkperformancedaily.com/2008/06/latency_and_jitter_1.html
posted by roboton666 at 4:47 PM on April 11, 2011
robot, that's just one of the factors which may (or may not!) be what the poster is asking about. Without more information, we're just playing pin-the-answer-on-the-network-stack.
posted by coriolisdave at 6:42 PM on April 11, 2011 [1 favorite]
posted by coriolisdave at 6:42 PM on April 11, 2011 [1 favorite]
Response by poster: Thanks for the responses!
I am only talking about network latency and apologise for the ambiguity and wooliness in what I have put. Ping world be one way of measuring latency as I suppose would be placing latency sensitive applications on the EC2 and comparing them with identically configured virtual machines on the LAN.
posted by conrad101 at 7:18 AM on April 13, 2011
I am only talking about network latency and apologise for the ambiguity and wooliness in what I have put. Ping world be one way of measuring latency as I suppose would be placing latency sensitive applications on the EC2 and comparing them with identically configured virtual machines on the LAN.
posted by conrad101 at 7:18 AM on April 13, 2011
This thread is closed to new comments.
posted by ripley_ at 3:08 PM on April 11, 2011