Hacker News new | ask | show | jobs
by obfuscate 5575 days ago
For a data set in the mere tens to hundreds of GB (in MongoDB, if anyone's curious), is there any reason I shouldn't conclude from this that I should use instance storage only (with multi-AZ replication and backups to S3, both of which I would be doing in any case)? Moderately slower recovery in the rare event of an instance failure seems better than the constant possibility of incurable killing performance degradation.

(Edit: I hadn't considered the possibility of somehow killing all my instances through human error. Ouch. That probably warrants one slave on EBS per AZ.)