Is instant recovery compatible with inline deduplication? That’s the question I found myself asking after listening to the briefing for Altaro VM Backup 7, since Altaro announced the addition of both features in this version. I’m a fan of instant recovery, which Altaro calls Boot from Backup, and I’m a fan of deduplication built into backup software. I’m curious to know if they can be effectively used together.
A Little About Altaro
First, let’s give credit to Altaro for getting me thinking about the subject. Altaro is releasing VM Backup 7, which adds inline global deduplication and boot from backup to their existing backup product which can protect Hyper-V and VMware environments. It seems to hit all the standard checkboxes. They offer three versions of the product, a free version, an Unlimited Version and an MSP version. Deduplication and boot from backup are available in the Unlimited and MSP editions.
Altaro Backup already had the ability to backup VMware and Hyper-V VMs while they are running, and to replicate those backups to an off-site location. They even support the idea of backing up to an off-site system that uses RDX removable disk devices that support drive rotation. They support file level backup in all their versions, and they support item-level restore of exchange in their enterprise and MSP editions. Again, the “news” is deduplication and boot from backup, which brings me back to my main thought – are these two features compatible?
Revisiting Instant Recovery
Instant recovery, or boot from backup, is one of the most important features to come out in backup and recovery products in the last decade. If your original storage was damaged in any way, that damage must be repaired before a traditional restore can begin. Once your storage is fully operational, you must then wait for the actual restore to take place before you can begin using the application again.
Instant recovery does not require the original storage to be operational for it to begin. All you need is a functional hypervisor to point to the backup system which will be used as the primary storage system during the recovery. You now boot your server from the backup, which allows you to begin using the application before you have even begun repairing the original failure.
IT can also use instant recovery to test backups on a regular basis. You can define a recovery group and make sure that all of those VMs boot up via instant recovery. If they boot, your backup works. If they don’t boot, you can figure out the problem before you need to actually fire the backup system in anger.
Now let’s talk about deduplication, which is the elimination of duplicate data from the backup image. It’s a very valuable feature that saves a lot of network bandwidth and storage, especially if it is done at the backup client the way Altaro is doing. It’s even more valuable if the deduplication is global, meaning that it identifies duplicate data between machines as well as between backups. Altaro tells us that their deduplication is indeed global.
Inline deduplication is where the data is deduplicated before it is ever written to the backup disk. This is in contrast to post process deduplication where the native backup is first written to disk, and then is de-duplicated as an asynchronous process. The advantage of inline, client-side deduplication is that it saves both network bandwidth and storage. The advantage of post process deduplication is that it may offer performance advantages when data is read from it randomly, such as what would happen during an instant recovery.
Instant Recovery vs. Deduplication
A lot of IT professionals are excited about the prospects that instant recovery brings to the table. They are also concerned about potential performance implications of various architectures. It’s easy to boot a single VM from backup, and if that is all you plan to do, then most instant recovery systems will probably be sufficient. The question is how well will the system perform if you have to recover several VMs at the same time using the same backup system for storage?
Conventional wisdom suggests that a post process architecture would have a performance advantage when booting multiple VMs, hence the question that started this blog post: is VM instant recovery compatible with inline deduplication? The answer, as always, is it depends. It depends on which deduplication product we are talking about, it depends on the hardware that the customer is using for their VM infrastructure, and it depends on what the customers recovery performance expectation is. In theory you can compensate for deduplicated performance problems by including faster disk and more processing power in the recovery appliance. The key to determining whether it is appropriate for you is testing to find the right balance.
At some point both inline deduplication and instant recovery will become commonplace and there will be less need to make sure that these two important features work well enough together. For now it’s wise to do a lot of upfront testing to make sure your expectations match the capabilities of the product. If you haven’t explored the idea of instant recovery, Altaro offers a 30-day trial for free.