Backblaze learned the hard way that data storage isn’t cheap. The company originally planned to build its service, which automatically creates a backup of a customer’s data to preserve in case the computer is damaged or stolen, on top of Amazon Web Services — or, when that failed, servers from HP or Dell — but couldn’t make the economics work in their favor.
So it did what any startup would do and built its own solution: the Storage Pod. And then it did what many startups wouldn’t. It turned down the many requests it received to sell the hardware it developed, refused venture capitalists who offered to invest if it pivoted from its original idea, and open-sourced the design so anyone could build his own Storage Pod.
That was in 2009. Today, Backblaze is releasing the design for the third generation Storage Pod. The new design “increases storage capacity to 180 Terabytes, adds anti-vibration drive bay assemblies, allows for the inclusion of multiple boot-drives including SSD drives, makes the chassis compatible with a wider variety of consumer drives, upgrades the motherboard and CPU and reduces time needed to install and replace any of the 45 drives.” Drop mic.
Backblaze CEO and co-founder Gleb Budman says the company doesn’t make “a single cent” off the Storage Pod, despite all the interest the server attracted when the Backblaze team first developed it. The founding team grappled with the decision to focus on its backup service, but Budman says that the company hasn’t considered entering the hardware business since the first Storage Pod design was made available online.
The Storage Pod, in conjunction with the decision to use consumer-grade hard drives for its storage, allows Backblaze to turn a profit despite charging just $5 per month for unlimited backups. Storage Pod pays for itself in this regard, allowing Backblaze to benefit from its development by enhancing the core backup service, not making a one-time profit from the hardware itself. By keeping costs low and developing its hardware and software together, Backblaze has built a system that currently stores 50 petabytes of customer data.
There is one other way Backblaze keeps costs down, a process Budman calls “hard drive shucking.”
You know how, when you want to eat some (delicious) corn on the cob, you have to pull the husk — or, as I like to call it, “that horrible green stuff in between me and some delicious grains” — away from the kernels? Backblaze did something similar with hard drives after a tsunami hit Thailand in 2011, wiping out global hard drive supplies and drastically raising the price of remaining stock.
“It was just craziness,” Budman says. Backblaze wasn’t able to purchase hard drives for its data servers, but the team noticed that external hard drives were still available at Costco and Best Buy. So the company bought a few, opened the packaging, removed the drive from its “shell,” and realized it’s the same drive used in the Storage Pod.
Backblaze was suddenly in the “drive-farming business,” as Budman calls it. The team mapped the location of each store and then purchased the retailers’ entire stock of external hard drives each day, repeating the process until Best Buy instituted a “two drives per customer” policy. The team kept at it until supplies came back up, obviating the need for the unorthodox procurement method.
Well, partially, anyway. Budman says that the team still cracks a few shells to get their hands on one of their preferred drives, a Hitachi-built hard drive that Budman calls “impossible” to purchase in any sort of quantity.
Those are the secrets to cheap data storage, it seems. There’s the Storage Pod, which Backblaze has made available (in design form, anyway) to everyone and is used by the Rensselaer Polytechnic Institute, the NASA Jet Propulsion Laboratory, and Heritage Auctions, among others. And then there’s “shucking,” which allows companies to acquire massive amounts of hard drives on the cheap, as long as they’re willing to get their hands dirty and crack a few shells.