Right now I am rsync'ing data from my laptop at home to an EBS volume attached to a m5.large. The write latencies I am seeing are ridiculously high (>15s).

My EC2 instance is EBS-optimized and read latencies look normal. The EBS volume is 300 GB and a gp2 (SSD) volume. My baseline IOPS for this configuration should be 900 IOPS and I am nowhere near that number right now. On top of that my burst credit balance is close to 100%.

Can anyone give insight on why I would see such bad performance?

  • 5
    How are you measuring latency? How did you determine that the bottleneck is the EBS volume and not bandwidth or network speed? – jordanm Mar 9 '18 at 21:05

The m5 and c5 instances are relatively new, and use a new method for mapping EBS volumes (NVMe). It has been plagued with problems, including the misreporting of disk latencies in CloudWatch. I do not recommend m5/c5 instances for production workloads at this time.






| improve this answer | |
  • Based on this answer the obvious thing to try changing to an m4.large instance to see if the situation improves. If it does you probably want to tell Amazon support so they can investigate the bug. – Tim Mar 11 '18 at 18:13

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.