Amazon EBS & Snapshot
WHAT IS IT?
We can’t talk about EBS without talking about snapshots, so we’ll cover both.
Amazon Elastic Block Store (EBS) is a virtual storage drive (like a hard disk or SSD) for your Amazon EC2 instances. You can think of it as the hard drive connected to your server.
WHAT FLAVOURS CAN I GET MINE IN?
1 : General-purpose SSDs balance price and performance for a wide variety of workloads.
2 : Provisioned IOPS SSD is the highest-performance SSD volume, best for mission-critical applications.
3 : Throughput optimized hard disk drive is a low-cost HDD volume designed for frequently accessed, throughput intensive operations.
4 : A cold hard disk drive is the lowest cost for less-frequently accessed operations.
5 : Magnetic drives are previous-generation drives, old and slow.
Anything your server writes to disk ends up stored in EBS as raw blobs of data called blocks. If you want to make a backup of that EBS volume, you take a snapshot of it and store it in S3. Each snapshot serves as an incremental backup for an EBS volume.
WHAT’S THE FUSS?
Like EC2 instances, EBS volumes scale vertically: if you outgrow your volume - needing to make it bigger - rather than splitting it into lots of little volumes. If you run out of space, it’s up to you to provision more storage for the volume; if your disk can’t keep up with the fast pace of writes to your busy database, you have up to the IOPS (input/output operations per second) as the limit. That requires monitoring and general operational awareness, plus it can get expensive.
It’s worth taking a moment as well to understand exactly what’s going on when you initiate a snapshot of an EBS volume. EBS snapshots are incremental: one snapshot contains the changes from the snapshot before it. And each incremental snapshot represents the incremental change from one volume to the next. If you have three snapshots and roll back to the second one, you’re deleting the difference in data between the second and third ones.
So if you were to say, “the third snapshot is my backup,” that isn’t necessarily true. Rolling back to that third snapshot will be reverting to the machine state at that time, but the third snapshot isn’t a backup of all of your data.
But snapshots do tend to save money as they only incrementally increase the data that you have backed up, rather than back the entire environment up over and over. If you’re regularly “backing up” your data, it makes more sense to take multiple snap shots rather than replicate the entire volume over and over again.
