Back several months ago we looked at how a company called Backblaze was finding creative ways to get HDDs through retail channels for their on-line storage business during the depths of the Thailand flood HDD shortages. The company would buy external storage boxes at lower prices than the available single hard disk drives and then “shuck” the external storage boxes to recover the drives that they then used in their own
Backblaze made their first generation “Storage Pod” design open source in 2009 and they have just announced the third generation of the Storage Pod (see http://blog.backblaze.com/category/storage-pod/). Prior Storage Pod designs have been built into many build-it-yourself storage systems and some commercial OEMs have also picked up on the design. The third generation increases storage capacity to 180 TB using 4 TB HDDs and allows multiple boot-drives, including SSD drives. This product is open source like the earlier designs and the estimated cost of the boxes is $1,950 without drives. With drives they estimate storage costs are about $0.06/GB, close to the naked HDD costs of about $0.04/GB.
Backblaze represents one approach to reduce the costs of digital storage and other enterprise hardware. The OpenStack Foundation and the Open Compute Project is a community of 150 companies committed to building cloud architecture based upon open source software and open hardware standards. The group includes Rackspace, Red Hat, AT&T, Cisco, HP and Dell. Seagate Technology announced that it joined this effort to help cloud builders develop more scalable, customizable solutions using open platforms while reducing operating costs. Storage device companies such as Seagate can provide cost effective storage solutions that could have special power saving or performance features that would be attractive to open storage system designers.
The Open Compute Project was the result of an activity at Facebook to build the most efficient computing infrastructures at the lowest possible cost. The Open Compute Project intends to develop servers and data centers that follow the model traditionally associated with open source software projects. On the storage side at the http://www.opencompute.org/projects/storage/ web page they list several current projects. Their cold storage specification is for a bulk load fast archive that handles a series of sequential writes and random reads. Such a system would be used to store data on HDDs that is almost never read, such as video surveillance or legal compliance data.
The Open Compute Project storage page also references Fusion-IO PCIe storage cards with up to 3.2 TB of storage capacity, which are used by Facebook to speed up overall storage system performance. The HYVE is a 2XOpenU storage server with up to 15 e.5-inch HDDs in a 3X5 array. This configuration is compatible with an Open Rack system to provide an off the shelf SAS expander. The Open Vault has a modular I/O topography for the Open Rack and holds 30 drives in a 2U chassis, operating with almost any host server. Open Rack is another Open Compute Project configuration. It is the first rack standard designed for data centers that integrate the rack into the data center infrastructure.
With the amount of digital data in data centers growing annually and storage budgets highly constrained open source storage system designs can fill an important niche for data centers and IT people comfortable with installing and maintaining their own hardware. Developments like those of Backblaze and the Open Compute Project are re-defining the possibilities of enterprise infrastructure and will likely find homes in more and more data centers.
