I neeed some help. We are getting a really read throughput (30MB/Sec on a 6TB volume) also if burst balance is at 67%. I just tried copying a 15GB file from this volume to a gp2 volume and this is the throughput.

I thought that if burst balance is above 0 I could have up to 250MB/Sec and a baseline performance of 72MB/Sec on this size on sc1 (according to https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/EBSVolumeTypes.html )

How can I troubleshoot what's going on?

Many thanks in advance.

  • Your question is a bit long winded and difficult to parse. Would the concise version be "We are getting 30MB/sec per second copying from a SC1 cold storage volume to a GP2 general purpose SSD volume, instead of the 250MB/sec advertised. The burst balance is 67% on both. How do we increase it?" – Tim Apr 30 '19 at 7:32
  • Suggest you edit your question to show a screenshot of CloudWatch throughput graphs and burst balance graphs so people can see them directly. Also please tell us the EC2 instance you're using. – Tim Apr 30 '19 at 7:34

It's impossible to say what the problem is with the information given, but one possible answer is you may be limited by your EC2 instance throughput, which is separate from the EBS maximum throughput. The real world throughput is documented here.

For example, a t2.micro has a baseline throughput of 0.06Gbps, which is 60Mbps, or 7.5MB/sec (less overheads I guess). t2.micro peak is about ten times that, 80MB/sec. The t3 series is better, with the same baseline but peak of 5Gbps.

Try a t3.2xl and see if the copy goes faster. If it does this is likely to be the reason.

| improve this answer | |
  • Hi Tim, thanks for the reply, instance type is m5.xlarge. Here is a screen of the source volume: ibb.co/w0Dcsq9 It started at 125MB/Sec and then dropped to 30-35MB/Sec until the transfer ended.... How can I check if I was capped by instsance limit? Thanks in advance! – Jb81 Apr 30 '19 at 14:43
  • The baseline instance performance is about 155MB/sec, so based on that it's probably not the instance limit. You could get an m5.4xl spot instance and try it to see if it goes faster, and compare it with a m5.large, that would probably settle whether it's instance limit or not. If it's not then I can think some more, but I'd prefer you do that test first. – Tim May 1 '19 at 8:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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