How do Virtual Machine Snapshots work in VMware
Piyush Chordia 2700021NT6 Visits (12630)
When you read the word snapshot what is the first thing that comes to your mind is , "A Photograph which preserves the best moments of your life". Technically snapshot is very much the same with the difference that it preserves the state of some digital resource. In VMware a disk "snapshot" is a copy of the VM's disk file (.vmdk) captured at a certain point in time. This snapshot preserves the disk file system and the files stored on it which can be of any type (including all the operating system files). So if something goes wrong with your virtual machine, you can restore it to a snapshot which was working previously.
One can also create snapshots for different vers
So how do snapshots really work ? There's just one Thumb Rule to VMware's snapshot technology : "Snapshots only store the differences between the current state and the original state". It follows the copy-on-write scheme with every subsequent disk access. Lets try to understand what that means ...
Consider that you have a text file with the word "COMPUTING" stored in it.This file is a sparse in nature : which means it spans across multiple blocks on the disk.Step 1 below demonstrates this scenario. The black lines indicate the links on to the stored data. For demonstration purpose lets consider that each block on disk has only one character.
As seen in Step 3 , we now take another snapshot after saving the word "COMPUTER" in the file. On making more changes after snapshot2 , similar process is followed to create snapshot2.vmdk file. The new changed word in the file is "CONTAINER". As a result now neither snapshot 1 nor the base file are written two, but are still referenced. Snapshot 2 will store the new changes as compared to snapshot 1. If you read the first level links of the green blocks from top to bottom , the word "CONTAINER" is read with the fact that only 5 letters are stored in snapshot2.
Below are the list of changes made to the file
Step1 - Base Disk = COMPUTING
Step2 - Snapshot1 = COMPUTER
Step3 - Snapshot2 = CONTAINER
From the above scenarios its clear that taking snapshots in VMware involves only writing the differences in files changed from the time of the snapshot, not the complete virtual machine disk. This mechanism is similar to taking diff and patch in Unix, but in a more sophisticated way that diffs on a binary level with the knowledge of how a VMFS ( Virtual Machine File System ) is structured.
Now we have a clear idea about the Copy-On-Write Protocol - Every time a block from the base disk or previous snapshot is changed , copy it to the current delta or snapshot file, which implies that when you perform a snapshot restoration, it only has to rewrite the sectors that were modified since you took the snapshot. As a result snapshot revert is also super fast.
But the Question is, What happens when you revert to an older snapshot? The VM software throws away the contents of snapshot2.vmdk and starts over with reading contents from snapshot1.vmdk. During this all the Blue links in Step3 are replaced with Green Links. ( this is a similar strategy followed in deleting a node from a linklist ... Pure programming stuff ). Note : Links to Snapshot1 and Base are not changed.
There are two more important aspects of Snapshot management : Discarding a snapshot and Merging a snapshot (reverting to a non-immediate parent snapshot). I'll discuss it in my future articles.
A KB article from VMware on Snapshot management.