As it happens, back in 2013, I wrote a blog post entitled, “Will we ever use flash drives for enterprise backup?” Interested parties can view my thinking at the time, which was not long after Pure Storage released its first FlashArray™. My conclusion was that despite challenges, over time flash technology would reach a cost/benefit value that would prove irresistible. It might have seemed daft to say that at the time, but that’s one prediction I got right.
Fast forward six years and we have Pure Storage FlashBlade™, a multi-purpose, high-density, all-flash system that’s perfect as a backup target. And it’s even better as a platform to drive recoveries and data re-use scenarios, which adds real value to your backup infrastructure.
The storage industry has been selling so-called Purpose-Built Backup Appliances (PBBAs) for decades. What makes FlashBlade so different?
First of all, the “Purpose” of the Purpose-Built appliances is backup, not recovery or anything else. They are designed to ingest data, and PBBAs do a pretty good job of that (though you’ll need to deploy several racks of gear to equal the ingest rate of a FlashBlade that fits into half a rack). PBBAs are also good at data reduction, though that’s often handled by the backup software instead. But one thing they aren’t good at – in fact, they are pretty terrible at – is data recovery.
When you design something specifically to ingest data, you make design sacrifices that result in slow data recovery. For instance, data deduplication techniques that spread blocks all over the place to reduce total data stored (data reduction being a Purpose). This results in loads of disk thrashing to re-assemble that data when it’s time to recover. If your recovery task is a few small files, it’s not going to be an issue. But if you’re restoring a volume of a million files or a fat, multi-TB database, you’ll have plenty of time to go for a coffee (and a donut, and a nap, and breakfast the next day) before your restore completes.
This may shock you if you’re used to the PBBA world, but FlashBlade actually restores faster than it ingests data, and it ingests quite fast. A half-rack sized FlashBlade deployment delivers 90 TB an hour of ingest (fast) and 270 TB an hour of restore speed (insanely fast). That’s not a typo for 2-point-7 TB. It’s two hundred and seventy TB an hour. Sorry if we ruin your donut break!
There’s another key aspect to keep in mind as well. Organizations increasingly seek to derive more value out of their backup data. The backup pool shouldn’t be a graveyard of dead data that only gets touched when you need to recover. It’s DATA, for goodness sake! And you should use it for lots of value-added stuff, like test-dev, reporting, training and insert-a-process-that-needs-data. The backup software vendors are very good at delivering the workflows to re-use data. You can spin up VMs from backups into a fenced environment, or instant-mount database backups for DBAs. But the problem here is that these features don’t scale with your typical PBBA. Spin up a few VMs, ok. Spin up a hundred? Or mount a few dozen databases for high-impact analytics? Your PBBA will grind to a halt. Why is that?
The typical PBBA these days does have some flash storage used for caching. You might refer to them as kinda-flash backup targets. Underneath the cache layer are good old spinning disks, often not even very fast ones. Once you flood the cache layer, your performance comes tumbling down. After all, re-using data is the same as restoring it to a target server for a specific use case. We already know PBBAs don’t do the restore thing so well. Oh, but FlashBlade does! It has no spinning disk to slow performance. It’s flash all the way down.
That huge restore speed of FlashBlade means it’s also really, really good at letting you re-use your valuable backup data to do lots of other important stuff. You can leverage all those cool features in your backup software at scale. And you don’t have to change anything about your backup environment other than the target device.
There are other neat design features in FlashBlade that make it a joy to behold. For instance, it simplifies data center cabling because it has an internal switch architecture. But we’ll let those go for today.
And I’m sure many are wondering, “Isn’t all-flash too expensive for backup?” The answer is no. We’ve got that nicked too, especially when you look at total cost of ownership. Though maybe you only need the power of FlashBlade for certain backup workloads that demand super-fast recovery SLAs and/or contain data you want to re-use. You have to start somewhere.
Since FlashBlade can also serve as fast primary file storage, it doesn’t meet the technical definition of a PBBA. But I like to think of FlashBlade as the Pure Better Backup Appliance, so I can still call it a PBBA. At least when nobody’s looking.
To learn more about FlashBlade, I recommend this lightboard video that explains how FlashBlade can make your backup environment better. This happens to include our backup partner Veeam, but Pure has many validated backup partners, including Commvault and Veritas. We also have white papers for using FlashBlade with native database backup tools, such as Oracle RMAN, Microsoft SQL Server and MySQL.
Going back to my 2013 blog post, I concluded with the following: “A few years from now it will be common to use flash arrays as backup targets, and probably a few years after that we’ll all wonder how we ever used anything else.” That time is now.