image_pdfimage_print

Using all-flash storage for backups is a subject I first wrote about 10 years ago in a blog post published on Computerworld. Back in 2013, it was an unprecedented idea, but using my crystal ball, I predicted the following: 

The way I see it, 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.

I was only slightly off on my timeline. Clearly, using flash for backups is no longer a radical idea, and those who have made the move are thrilled with it. But most backups still land on spinning-disk or hybrid systems. It’s time to shake off that old way of thinking because at this point, using disk for backup just doesn’t make sense. 

Most backups still go to so-called purpose-built backup appliances (PBBAs), such as PowerProtect DD (formerly Data Domain), HPE StoreOnce, and dozens of others. The “purpose” of purpose-built appliances is backup. They’re designed to ingest data, and PBBAs do a pretty good job of that. PBBAs are also good at data reduction, though that’s often handled by the backup software instead. That’s the good part. What about the downsides?  

The problems of PBBAs stem from two areas. One is a software design issue. The other is simply what they’re made from. 

Ransomware

The Trouble with PBBAs 

When you design something specifically to ingest data, you make software sacrifices that result in slow data recovery. PBBAs use deduplication techniques that spread stored data blocks all over the place. 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 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. If you’re recovering your entire data center because of a ransomware attack, you could be looking at days, weeks, or months to recover—and in some cases, never. 

When you use Pure Storage® FlashBlade//S™, you don’t have this issue because there’s no seek time with flash. 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. With the proper network and backup software infrastructure, you can drive hundreds of TBs of recovery per hour. Sorry if we ruin your donut break!

But not all data is created equal, and there are tiers of data importance, which is why Pure also offers FlashBlade//E™, a capacity-optimized array that’s perfect for less critical data or for longer-term retention. And FlashBlade//E sells at approximately the cost per GB of disk, but with much lower operating costs. Most modern backup software can easily tier backups across storage targets, so a two-tiered strategy makes sense. 

The Last Days of Disk? 

The other problem with disk-based appliances is, well, they’re based on disks! What’s wrong with that? Here’s a list of issues with disk: 

  • Power consumption. Disks are power-hungry beasts, and when you have racks and racks of PBBAs, you can literally feel the power being consumed (walk through the hot side of the aisle). But most organizations are looking to cut back on power and emissions, and in many cases, data centers simply can’t bring in any more electricity. By switching to flash, you can cut power by 80% or more. That’s a really big number, and replacing backup appliances is a great way to get greener in a hurry.
  • Racks and racks and racks. And more racks. Even though PBBAs were designed to save on space, over time, massive data growth forced many organizations to have dozens or even hundreds of them deployed. Data center space isn’t free, and as space-hungry initiatives like AI and analytics grow, where are space savings going to come from? Easy. Replace your PBBAs with much denser FlashBlade systems.

We recently sized a replacement at a major manufacturer, and they went from 45 racks of a popular PBBA to 4.2 racks of FlashBlade. They’ve got 40 empty racks now. That’s a 91% space savings. Space problem solved. As a side benefit, fewer appliances means fewer network ports, which means even more cost and management savings. 

  • If it spins, it breaks. Disks may seem reliable, but compared to flash, they’re terrible. And as the number of disks increases, simple math makes failures more frequent. We’ve seen many large enterprises that have “disk replacement days” each week when the IT team focuses on replacing all the disks that have failed. That’s a lot of effort to basically stand still. FlashBlade is on average 20 times more reliable than disk systems. You’ve got better things to do than spend time replacing failed hard drives. 
  • Forklift upgrades and e-waste. If you’ve been using PBBAs for any length of time, you’ve probably seen quite a few of them hauled out of your data center and basically dumped into a landfill. As organizations are striving to reduce e-waste, be aware that Evergreen® technology from Pure means arrays can be expanded or upgraded in place without a full system replacement, leading to an 85% reduction in e-waste. You’re also saving the trouble of migrating data and reconfiguring backup jobs to new appliances all the time.  

There’s a lot more goodness about FlashBlade, including built-in ransomware protection, tested compatibility with multiple backup vendors and database tools, and support for both file and object protocols. (My infallible crystal ball tells me that object storage is the future of backup, even for high-performance use cases). 

While you may not be able to replace all your PBBAs at once, as they come up for maintenance or lease renewal, you owe it to yourself and your organization to take a good long look at replacing them with all-flash. The benefits are undeniable. Give it a try, and I’m sure you’ll agree that disk is done when it comes to backups. 

Get the technical details on FlashBlade//E and FlashBlade//S