BlackSky With Scalable Storage Architecture for HPC

BlackSky Computing, providing HPC and storage system solutions purpose-built by users for users, unveiled their new Apollo enterprise storage system for HPC at SIGGRAPH 2011.

The performance and massively scalable Apollo enterprise storage system for HPC has a small, green footprint packing 180 TB in a 4U chassis which can be expanded to support a single file system of over 2 Petabytes with no degradation in performance. The BlackSky storage system can work with HPC data sets up to a 80 Gb/s resulting in a storage system that has four times faster the I/O and one fourth the price of the leading competitors.  

"We created this product because we looked for an HPC storage system for our own internal cloud and for our customers for more than a year and found nothing that satisfied all the requirements for compute intensive applications," said Scott Alexander, co-founder and CEO, BlackSky Computing. "The only offerings with sufficient I/O were cost prohibitive because they weren’t designed for high performance computing or render farms. The Apollo storage system was purpose built to deliver the three essential features for HPC storage – tremendous capacity at low $/TB; proportional network bandwidth in and out of the device which will cripple HPC applications if not done correctly; and expandability under one file system to very large volumes."

The Apollo enterprise storage system is a redundant system, including two mother boards, and supports RAID 1 and 6, ensuring that if any drive fails it can be replaced without loss of data. The system also includes a web-based user interface that simplifies initial installation as well as future expansion.

"We’ve seen situations both on our cloud and in customer environments in which the HPC systems were operating at just several % capacity for long periods of time because they were waiting for data from the storage array. This is where many existing solutions fail and where the BlackSky Apollo storage system will dramatically increase the ROI of the HPC environment," added Mr. Alexander.

The Apollo enterprise storage system joins the Hyperion Enterprise HPC system in the company’s purpose-built product suite. Hyperion offers blade server functionality at traditional server pricing and is able to use the latest, fastest Intel processors. The dense Hyperion design uses half the space of traditional 1U servers and is 2 to 4 times more dense than the competition, has highly efficient power supplies, and provides for plug-and-play expansion. A standard 5U rack mount chassis can support up to 10 dual Intel Xeon X5600 series compute blades, up to 20 CPUs, 12 cores per blade (120 cores total), up to 96 GB DDR3 1333 ECC RAM per blade (960 GB total), and up to 40 Gb/s I/O per blade. Moreover, every compute blade comes with its own network adapter and power supply. If any single component fails, it affects only a single blade. Failed compute blades can be removed, repaired, and replaced without having to power down any other running blades. Together, the Hyperion and Apollo offer performance, expandability and cost-effectiveness when combined in HPC applications that require massive computing power like rendering for visual effects and CGI, or calculating large data sets for research and development.

"We will continue to drive customer-centric innovation while further establishing our technological leadership to develop enabling high performance computing solutions to better serve our customers," concluded Mr. Alexander.