EBS Optimized instances 30 minute burst limit

Question

Customer would like to know how EBS optimized instances and the 30 minute of maximum burst works.

My guess is that the 30 minute is not a hard limit number but a good rule of thumb we gave our customers, but starting from ( https://docs.aws.amazon.com/AWSEC2/latest/WindowsGuide/ebs-optimized.html ), let's make an example:

if I have a r5.2xlarge, which has a baseline of 2300 Mbps bandwidth, and I use 3000 Mbps for 20 seconds, do I have 2 minutes and 40 seconds left to "surpass" my baseline or the 30 minutes is the number of minutes available if I use the maximum bandwidth? (e.g. if I were to consume ( (4750+2300)/2 = 3525 Mbps ) I would have 6 minutes?

Another related question would be: am I consuming "by the second" or "rounded to the minute"? 
e.g. if I use 20 seconds at maximum bandwidth, I consume 1 minute.

Accepted Answer

This blog goes more in detail on the EBS-Optimized burst behavior https://aws.amazon.com/blogs/compute/improving-application-performance-and-reducing-costs-with-amazon-ebs-optimized-instance-burst-capability/. We have burst balance metrics which help customers evaluate whether they can size their workloads based on burst performance.

Customer can drive sustained burst performance for 30 minutes, after which the performance will drop to the baseline. For example on the r5.2xlarge customer can drive sustained 4750Mbps for 30 minutes.  Most customer applications don't drive sustained workloads and such applications would always have burst performance available. I would recommend customers test their workload and monitor the instance burst balance metrics to determine whether  it meets their application demand.

EBS Optimized instances 30 minute burst limit

相關內容