I’m by no means a fully-fledged greybeard, but I’ve been in the storage industry for a while now. Long enough to see the vendors get excited about the death of tape some fifteen years ago. Tape may not be dead, but the idea of backing up to (and recovering from) spinning hard disks has certainly left a mark in data protection circles. A number of vendors have hardware appliances comprised of large capacity disks that can act as targets for backup data.
Secondary Flash Storage?
Generally speaking, the technology works well and provides a reasonably performing solution for most enterprise environments. The specification of the disk target will, in concert with the software that you use, have a great impact on the speed with which you can recover your data. The speed of these devices hasn’t been a huge problem in enterprise shops where they often have the luxury of using different systems for disaster recovery and periodic data protection. A number of places don’t have that option though and have found out the hard way that the recovery time on some of these appliances can be pretty long.
It seems silly to propose an All-Flash solution as a backup target. The industry still hasn’t quite reached the point, fiscally speaking, where it makes sense to deploy hundreds of terabytes of Flash storage for use as secondary storage. Even with deduplication and compression, you’re looking at some pretty big numbers to replace your spinning rust-based secondary solution with an all-Flash equivalent. But I’m not suggesting that you go out and buy a FlashArray to use for secondary storage. Right now, that doesn’t make sense commercially.
Keeping Your Apps Available
Nonetheless enterprise shops and web scale organizations alike are faced with a very real requirement to look at data protection in a different way. There’s an incredible amount of pressure to have applications available 24/7, and more and more companies are talking to me about the expectation that their stakeholders have that data can be recovered in an instant. You can protect your infrastructure with something like ActiveCluster, but that kind of approach might be overkill if you only have one or two applications that have that kind of super critical recovery point objective. Or you might have applications that don’t necessarily respond well to crash-consistent replication and need a more delicate approach taken to protecting them.
One solution has been to adopt a per-rack approach to data protection, with servers being loaded up with direct-attached storage. The applications are then backed up using application-specific tools and copied to a host in another rack. The idea is that you’re protected because you have more than one copy of the data in play, and recovery is quick because you don’t have to go back to some central repository to retrieve the data. Nor do your applications people need to talk to those infrastructure people. Infrastructure people can be so annoying, always talking about patching hosts, and alerts, and rack space, and capacity. Applications folks just want to get stuff done, right? In any case, if you’ve adopted this approach to data protection, you’ve probably enjoyed some of the advantages such as quick recovery and independence from obstreperous infrastructure staff. But now your applications staff are becoming capacity managers and having to devote time to managing space usage in their servers, more than they have previously. And for every backup version you’ve kept on the server, there’s possibly only a small amount of disk space saved though native compression.
FlashBlade Cuts In
Imagine if you could take those couple hundreds of gigabytes worth of capacity from each host and put it in a dedicated appliance, that you could in turn back up your applications to? It’s accessible via NFS, it’s got some networking chops, it’s all-flash, and there’re all kinds of data reduction going on. And while you’re at it, why not use the FlashBlades for other cool stuff, like analytics and business intelligence? Fast backup is great, and Pure’s Rapid Restore feature is even better. You can leverage native tools to perform your data protection, or you can leverage a number of leading partners to integrate into more traditional, enterprise protection if required. Your teams can recover data almost instantly, as required, without having to deal with complex silos of storage and processes.
Your applications folks are happy, because they can still recover their data quickly and they don’t need to be systems administrators. And they have the added bonus of having a data platform they can use to do more cool stuff with their data. Your infrastructure folks are happy, because they can keep an eye on what’s actually being protected. And they have the added bonus of not having to deal with racks and racks of poorly utilised secondary storage. That’s why FlashBlade can make sense.