Jason Nash’s Blog

The great thing about virtualization is that it makes servers so easy to provision.  On the other hand, if you’re the guy doing backups is that a good thing?  Unrestricted, servers can multiply like little bunnies leading to a huge amount of growth in data to retain and backup.  Plus, you now have more servers on less hardware so when the backup window hits you’re pushing more data through the same pipes and using the same CPUs.  Now what?

This problem isn’t just in virtualization.  The data that most companies must manage is growing at an amazing rate.  At some point you just can’t push all of that to tape or other traditional media.  Most organizations are finding that they must figure out new ways to backup and retain all of this information.  The first step is usually archiving that data.  The second is to figure out how to backup and retain data more efficiently and effectively.    We’ve talked about Avamar several times on this blog before, and for good reason.  Avamar helps solve this problem by using data deduplication.  Unlike other solutions it does this on the client side so only the unique data that the Avamar datastore does not have gets sent over the network to be backed up and stored.  Efficient!  How can we leverage this with VMware?

Avamar works very well with VMware.  That shouldn’t be a surprise.  EMC owns VMware, right?  But what are our options?

Agent Based Backups

We’ve all done this.  It’s the standard method for backing up servers no matter what backup app you use.  You install a backup agent on the server.  In a VM environment you just have one installed on every VM.

The Good

It’s simple.  It works.  Nothing else required.  The implementation couldn’t be easier.  You get all the benefits of the Avamar de-dupe technology without worrying about anything else.  So, we’re done!  Right?

The Bad

Well, not so fast.  Yes….it’s what we’ve been doing for a while.  But is it what we want to do in a virtual environment?  Maybe not.  Why run 10 agents on the same ESX host if you don’t have to?  Agents also limit your restore and recovery options.  You’re basically going back to the old ways to do a full server recovery.  Yes, VMware makes spinning up a server real easy but you’ll still spend a lot of time getting it just right so you can restore your apps and data.  Also, while Avamar doesn’t use much CPU on those nightly backups thanks to the really good data set reduction, you still can do better.

VMware Consolidated Backup

Avamar supports VCB.  Some people really like VCB.  Others do not.  I think a lot of the dislike of VCB is tied to people using bad or immature tools so don’t be too prejudiced here.  Like other VCB solutions everything is done through the VCB Proxy server.  In this case you install the Avamar agent on to the VCB Proxy.  This server has access to the snapshots of your datastore LUNs and just backs up the image files directly.  See the diagram below:

The Good

This methodology is similar to other VCB-integrated solutions.  The benefits include no agents in the VMs and if you architect it correctly you can do your backups without affecting performance on the production networks.  The VCB integration also allows individual file restores from the VMDK files.  Avamar still does a great job of deduplication on the VMDK files, just like it does on the files inside it with an agent-based setup.

The Bad

Not a lot of things wrong with this option.  There is an extra cost associated since you have to license VCB.  Configuration of VCB can be a problem, but the benefits are well worth it.  This is probably the best solution as you can do image based backups for fast restore but still have the option for single file restore.

Service Console

The final option is to install the Avamar agent directly on the ESX server’s service console.  Since ESXi doesn’t have the service console this is not an option for those users.  This is an interesting idea and really only useful in small shops without a SAN or other consolidated storage.

The Good

It’s easy since you just install the Avamar agent in the console one time on each ESX server.  No VCB license required and you can do it with any storage type.

The Bad

It uses the processing and connectivity of the ESX server.  Also, since it’s backing up the VMDK files directly it can’t really back up an open VM.  To get the best backup you must suspend or power down the VM.  Also, since you aren’t using VCB you can’t access the VMDK file and restore an individual file.  Clearly not the best option for most people.

Conclusion

As you can see there are several good options for integrating Avamar in to your VMware environment.  Avamar really shines in a virtualized environment.  It greatly reduces the amount of data you backup and store while using your system’s resources very efficiently.