[falcolas]

> Stripe your RDS disks for better performance

This is a fun hack to perform, but it opens you up to another problem: latency on any of the striped EBS volumes will lag out the entire striped array.

Attempts to mitigate this problem (including setting up raid 10) work in the short term, but it really is easier to just purchase a guaranteed iops volume if you want to run a database on EC2.

[DrJ]

RDS is the hosted sql database solution.

When RDS stripes your disks (I have seen it start when I jumped from 100GB to 300GB) you benefit from faster IOPS (I have seen 700 IOPS improve to 2500 IOPS.

PIOPS (Provisioned IOPS) is the way to go in the long run though.

[falcolas]

> you benefit from faster IOPS

Yes, you do.

However, if a single EBS that backs the RDS instance lags out (as EBS volumes in AWS are wont to do), you lose the iops.

The key point is that as the number of EBS volumes you depend on goes up, the chances of this happening goes up correspondingly.

Striping EBS volumes works, but I would not depend on it for a production environment.

[rpedela]

Yeah that is one reason we just use instance storage with database replication. Instance storage is generally faster and cheaper than EBS especially if you have a "very high I/o" instance.