AWS Storage gateway sizing

[veehexx]

We're in the process of implementing the AWS Glacier tape VTL system into veeam. currently running the minimum 150GB for both cache and buffer.
As a test backup, 90GB backup job at 71% job completion, AWS is reporting 15% buffer usage.

We've also had some failed test jobs (CRC errors) in the last week which are 7TB+ datasets so based purely on the above 90gb test i can forsee 100% buffer usage assuming the usage is linear. CRC failures i believe I've resolved now.

for those who've production experience with AWS VTL backups, how have you come to determine the cache&buffer size requirements? i cant find anything online about correctly sizing them, other than the min/max limits Amazon provide.
what happens if your incorrectly sized - performance issues only, does job reliability become an issue, does it just limit backup job concurrency, or something else i've not considered?

[Dima P.]

Hi veehexx,

From the top of my head the only issue was discussed here and it was related to the cache size limit being reached.

what happens if your incorrectly sized - performance issues only, does job reliability become an issue, does it just limit backup job concurrency, or something else i've not considered?

If I remember correctly the entire tape should be stored in cache prior upload to Amazon, so you have to estimate how many concurrent tapes are going to be used at the same time and then allocate cache based on the emulated media size. Once it's uploaded cache can be used by other emulated tapes. Cheers!