Restoring Entire Virtual Machines in 6 Seconds with Rubrik's Instant Recovery Technologies
Data protection is an insurance policy used to guard against loss, corruption, and failure. Because of this, I’ve often said that the true measure of any backup solution is the ability to restore data. After all, backups that can’t be restored aren’t really worth that much to you and your team, right?
There are a number of quantifiable methods for putting your eyeball on the ease of a restore, with search time, recovery time, and overall simplicity of the workflow coming to mind. The amount of moving parts involved to restore data have a direct and exponential effect on restore times. Because most all solutions require combining disparate storage and software together, many of the top solutions in the Gartner Magic Quadrant require 10 – 30 minutes to restore pieces of data. This is measure from the start of the request to the end of the workflow.
That seems like an incredibly long time and there are a number of reasons for it. Other solutions have to hydrate data, transfer it from one media to another, and are using write-optimized storage volumes for restores (this doesn’t work very well). Enter Rubrik: a converged data management stack that uses a combination of flash and capacity drives, along with a very clever file system named Atlas, to deliver tens of thousands of IOPS in each 2U appliance.
Mounting a Copy
Below, I’ve created a short video demonstrating our Mount feature, which allows you to restore a copy of an entire virtual machine from a backup point in just a few seconds. This particular VM has 40 GB of thick provisioned disk and is running Windows Server 2012 R2 – it’s actually my lab virtual machine that I use to run various tests. Notice that once I kick off the Mount request from Rubrik, a full copy of the VM begins building and powering on in the vSphere Client in a matter of seconds. Even accounting for API overhead between vCenter and Rubrik, the VM is constructed in a mere 6 seconds and then begins to boot up.
Did you notice that it only takes two clicks to perform an Instant Mount? All you need is a date selection and ESXi host selection. The remaining complexity is abstracted away by Rubrik and handled on my behalf. The entire workflow is also easily available from the RESTful API or PowerShell Module at no additional cost (all features are licensed out-of-the-box).
This is a handy workflow for:
- Building a copy of a virtual machine for pre-production environments
- Instantiating a complex, multi-tiered application within an isolated network for upgrade or regression testing
- Performing analytics on a copy of any structured data set
- Cloning environments based on rules and triggers dictated by the application / development lifecycle
Recovering from a Disaster
If, however, a disaster has struck, there’s another option based on similar mechanics – the Instant Recovery. This is the killer application baked into Rubrik natively, and it allows for a full virtual machine to be brought online in seconds while also deprecating and replacing the original, faulty VM.
I’ve selected my CentOS virtual machine and have simulated a disaster by completely removing its VMDK. There’s no longer any data associated with the VM, and it cannot boot without the required partitions. Using Rubrik, I use the predictive search feature to find my VM, then a two click restore – the backup data point, and the ESXi host target – to return to operations. As with the Mount process, Rubrik has abstracted all of the underlying complexity associated with:
- Building a flash-fueled datastore and mounting it over NFS
- Presenting the VM files based on the backup date stamp
- Marking the original VM as Deprecated and powering it off
- Powering on the backup copy using the original name
Here’s a quick video demonstrating the Instant Recovery process.
One of my favorite workflows to use with Instant Recovery is the idea of having a near-zero rollback feature for any applications I’m working with. I can bake in the SLA Domain’s policy, overlay it on a workload, and then know – with confidence – that the virtual machine can be brought back online, without any impact on my production SAN, in a handful of seconds. This technology really takes data protection to a whole new level.