I thought I would return to the subject of Avamar for a quick discussion of one of my favorite pieces of backup software. The geek in me loves new technology and new gadgets, and the contrarian in me loves the thought of doing things different–so long as different is also better.
Avamar meets both of those criteria.
I have written about Avamar for VMware before, and how it is such a good thing. And six months after writing that, it is still a good thing. It doesn’t matter if you opt to deploy it in a guest VM, or along side VCB proxy backup, it makes backup of a VMware environment vastly better than it is with traditional backup.
So call it an unintended consequence, but there is something else that happens when you elect to back up an ESX server with Avamar. That is, something else in addition to a reduction in the stress levels of backup administration roughly equivalent to a 6 month stay on Bora Bora, a reduction in storage requirements equal to a mountain of tape, and a reduction in bandwidth requirements of equally gargantuan proportions.
What happens? You get to consolidate more.
Avamar with VMware is so efficient at backup that you can fit more VMs on a single ESX box. When you reduce CPU utilization by 80%, and network utilization by 99%, your ESX server just got a lot less busy. And you know what we do with idle ESX boxes? We consolidate more.
VMware is all about consolidation, and this just lets you push it that much further.
Real world experience shows that if you had to stop consolidating at 10 VMs per ESX server with a traditional backup solution, you can push it to 12-13 VMs per ESX server with Avamar.
Let me put that another way: Avamar makes your VMware infrastructure 20-30% more cost effective.
That savings is equivalent to making Avamar free.
I like that math.
For a graphical look at what I am talking about here, consider the following two graphs, each of which depicts a backup job:
In this case, I am looking at what happens during the backup of a data set. My first graph shows CPU utilization during the backup of a single VM; on the left (in red) is with a “traditional” backup application, on the right (in blue) is with Avamar. Similarly, with network utilization, you can see the “normal” backup in red, and the Avamar in blue. Or rather, you really can’t see Avamar, because the reduction is so enormous, it becomes almost invisible.
Now stack a bunch of these VMs onto a single ESX system. You get contention–and contention prevents further consolidation. More accurately: you get contention if you are not using Avamar.
The real question becomes: can you afford not to use Avamar for VMware backup?