Backblaze open sources 135TB storage architecture

Less than two years after open-sourcing the design of the 67 terabyte “storage pods” that underpin its cloud backup service, Backblaze is at it again. This morning, it shared on its blog specs for a new, 135 terabyte storage pod that costs only $7,384. Its open-source designs have been wildly popular since Backblaze released them, but don’t expect it to get into the storage-vendor business any time soon.

Om detailed the rationale behind the move toward open-source storage when Backblaze released its first designs, which boil down to Amazon S3’s (s amzn) and traditional storage systems’ being too expensive. After almost two years of running its custom designs internally, Backblaze co-founder and CEO Gleb Budman told me, the company decided its true intellectual property was in the desktop client and back-end cloud software that power its service, so it thought it would share the love with regard to storage.

Not surprisingly (although surprisingly to Backblaze), its designs got a lot of attention. It turns out there are other organizations out there that could really benefit from high-capacity storage at $117 per terabyte (note: the new version slashes that price to about $55). Budman told me the company that sells the cases Backblaze uses changed its home page to read “We sell Backblaze storage cases,” and the manufacturer of a rubber anti-vibration component created a special link for the part because the company was inundated with phone calls asking how to buy it.

Backblaze didn’t escape attention either. Budman said about 200 companies contacted Backblaze asking to buy storage pods, and at one point the management team actually considered taking them up on the offer. Ultimately though, he said, Backblaze didn’t get into the business of selling storage because “it’s fundamentally a polar opposite business to the one we’re currently in.” Selling massive storage appliances to large companies is a far cry from providing a software-based backup service to SMBs and individuals.

Still, a large number of organizations have built their own Backblaze storage pods, including Vanderbilt University, which uses it for medical-image archiving, and at least one big advertising agency. And because everything is open-source, Budman noted, anyone who wants to start their own storage business based on the design is free to do so.

The open-source nature also paid dividends to Backblaze, which has created a development community of sorts. One well-known storage-pod user, Shutterfly, shared with Backblaze its strategy for improving performance by going with a new motherboard containing three PCIe slots and using better PCIe SATA cards. Backblaze’s first-generation design contained four PCI slots, but the system included a single standard PCI slot that created a bottleneck.

Other improvements in the latest iteration include twice the storage capacity because it now uses 3TB hard drives, and twice the speed because of the PCIe upgrade as well as more RAM and better CPUs. It all results in a 4U appliance containing 135 terabytes and performing twice as fast as the first generation storage pod for about $500 less than the first generation (which only housed 67 terabytes). Budman explained that Backblaze’s design makes it possible to fit a petabyte of storage into three-fourths of a standard rack, something that would have taken an entire row just a few years ago.

After seeing how popular its first-generation storage pod design was, Budman is excited about the new uses organizations might find for the bigger, better version 2.0. In combination with open-source server designs from Facebook, he explained, innovative software companies can almost guarantee themselves they’ll have the hardware tools to meet their needs. Gone are the days of spending a million dollars on equipment just to launch a website.

“Innovation is driven by having access to things,” Budman said. Now, if hardware vendors or cloud providers can’t meet a company’s needs around cost, they can just build their own components using readily available — and proven — designs.