In the previous blog, we discussed the good, bad, and ugly of EBS snapshots. In this blog we will elaborate why recoveries from EBS snapshots takes a very long time.
Here is how recoveries from EBS snapshots work.
When you mount an EBS snapshot to an EC2 compute, the EBS snapshot makes you believe that the data is instantly accessible, but is not available in EBS storage. As you or your application starts reading blocks of data, those blocks are pulled from S3 storage to EBS storage and then the data is “really” available. Of course all of this is happening seamlessly in the background.
So as long as your application can survive the delay in accessing data because of the hop from S3 to EBS, you are fine.
However most databases don’t like such behavior. In many cases, the database might not even start !!!
Let’s compare the above behavior with how on-premises storage snapshots work during recoveries. Remember, the production and snapshots were stored on the same storage. Thus, when you mount snapshots, the data is just readily available and there is no data transfer from a slow tier to a fast tier storage. Thus recovery performance off an on-premises storage snapshot is just fine with no delays as you saw above in EBS snapshots.
Of course the problem with on-premises storage snapshots is that if the production storage goes down, you lose the production and the snapshots and you have nothing to recover from.
Is there an ideal solution?
An ideal solution would give you all the best options such as:
- Protect data in an application consistent and incremental forever manner
- Store the backup data on a different EBS backup volume so that if you lose production, you still have the application data on the EBS backup volume.
- Store the backup data on object storage as well for long term retention to reduce costs
- Mount instantly from EBS backup volume so you get great performance with low RTO
- Mount instantly from object storage and access data quickly from points in time that are months, years, or decades old.
If you want to learn how this solution works, download the white paper to learn how Copy Data Management delivers low RTO, low RPO, and 74% cloud infrastructure cost savings.