Here's what existing Amazon EC2 performance benchmarks tell us -- 1)
EC2 is fairly variable in terms of absolute performance, 2) is within
the same ballpark as most commodity equipment and 3) EC2’s EBS
performs slightly less well than native disks.
The problem is that all these Amazon EC2 benchmarks are concerned with
raw performance. More useful EC2 metrics should focus on aggregate
performance or scalability. For example, a useful suite of benchmarks
would be analysis around metrics like “cost-to-grow per request.”
Read more about the limitations of existing benchmarks and the
importance of pursuing a different set of benchmarks that present a
coherent view of what the Cloud means in terms of web applications.
http://bit.ly/caNWb