Reply to post: Tangled storage

Storage on AWS: What's new, is it too complicated? Can it help defend against ransomware?

Kevin McMurtrie Silver badge
Mushroom

Tangled storage

The EBS options are maddening. There are multiple EBS description pages of varying obsolescence and completeness. Once you find one that looks current, you can spend a good 4 hours trying to figure out which ones to use. GP2 is the cursed storage with a burst balance and throughput that varies by size. GP3 is better but not always documented. Then there are ephemeral devices, which are allocated like EBS but not EBS. If that's not all crazy enough, EBS is assigned to instances using a device identifier that has no relationship to the actual device path. The actual device path varies by instance type and storage type so your init scripts can do little better than guessing where everything is.

So you want to start up your instance with lots of data? Amazon recommends EBS snapshots. Just map EBS device to a snapshot and... and... and... and you're still waiting for processes to launch. EBS snapshots perform at 4 to 16 MB/sec until converted into something more local, but not actually local. You don't need to scale up with less than a day's notice, do you? AMIs aren't a workaround because they're snapshots too. You have to subscribe to Fast Snapshot Restore to make it faster. Exactly how fast? Faster. Think you'll pull in data quickly from another service? You'll have to allocate the IOPS for that and pay for it for the life of the volume.

Oh, and can the S3 API please support streaming upload so temp files aren't needed to calculate the final size? You know, because the /tmp directory is an EBS volume.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2022