ec2 restore performance with and without a proxy

[mrt]

I'm seeing faster restore times when an EC2 proxy is NOT used.
The repository is hosted on EBS.

restore without a proxy:
copy to temp s3 bucket: 2.5 minutes, 25GB @ 374 MB/sec
import process: just over 23 minutes
about 26 minutes total

restore with a proxy (4 cores, 16GB ram)
restore vmdk process: about 5 minutes (25GB @ 128 MB/sec)
snapshot process: 14 minutes
import process: 16 minutes
about 35 minutes total

Is this expected? Does 35 minutes to process a 25GB vm seem reasonable for a proxy? I was hoping a proxy would improve performance. Thanks

[HannesK]

Hello,
looks expected to me for the amount of data, the overhead of the proxy and the bandwidth available. The key problem is the import API by AWS. It surprises me, that you had such significant differences, because the API is always the same. But that depends on AWS infrastructure.

Note: I recommend to check whether you have an orphaned snapshot for the restore test with the proxy. There is a known bug (we plan to fix it) that snapshots might not be deleted correctly.

Best regards,
Hannes