Terascala, Meet LSI. Team, Meet HPC.

Print Friendly, PDF & Email

Last week, we covered the press release from Terascala detailing new product lines and partnerships featuring the disk sub-systems of LSI. File system appliance startup meets seasoned pro in enterprise storage. However, most fail to realize the potential significance of such a partnership. Positioned correctly, the joint efforts of Terascala and LSI could realistically change the game of HPC storage. We had the pleasure of speaking with Rick Friedman, Terascala’s VP of Marketing and Product Management, and Steve Hochberg, LSI’s Director of HPC, to find out just how.

The story begins with the somewhat ubiquitous parallel file system, Lustre. Born out of funding from the Department of Energy, it serves as the basis for quite a number of the largest supercomputers in the world. It has proven time and time again that we can successfully scale file systems to meet the immediate data demands of today’s petascale computing environments. This, however, has come at a price. As with many other technologies on the lunatic fringe, Lustre requires a certain expertise in order to maintain stability, reliability and performance. In today’s HPC talent market, this expertise is a rare commodity.

Enter Terascala, stage right. In the midst of the data scalability explosion, Terascala took on the ominous task of producing a reliable, scalable, high performance Lustre appliance designed to meet the requirements of the next generation of computational procurement. The company founders saw that HPC users and administrators were changing their attitudes towards file systems. File systems were no longer “accessories” to machine procurements, but rather a part of the central requirements. As such, the core goal of Terascala was founded in providing an appliance to the HPC masses in order to eliminate the systems management overhead associated with large-scale parallel file systems.

Terascala began their journey producing their Lustre appliances packaged with basic features and relatively small home-grown disk arrays. They quickly realized that customers would require more features and enterprise-class disk sub systems. They went on a search to find an appropriate disk subsystem provider that met the enterprise reliability and scalability needs found in the latest HPC procurements. According to Rick Friedman, LSI had the technology, the expertise and the gumption that Terascala was looking for in an enterprise partner. “[LSI] really understands what the HPC environment is all about, not just what’s on the spec sheet.”

The first product of the new partnership is the Terascala DTS 4000. It combines the latest series of redundant Terascala Lustre appliances with the LSI 7900 HPC XBB2-based storage system. What results is a file system and disk solution that can be configured for a mixture of reliability, capacity and performance in a variety of ways. Customers can get preconfigured Lustre systems with up 480 drives per object storage module and metadata systems that support up to 15,000 file creations per second. As a reference, Terascala provides customers with details for a 960TB configuration based on the DTS 4000 with included redundancy.

What has resulted is a partnership juggernaut that takes the guesswork out of deploying and managing super-scale parallel storage systems. No more are the storage system administrators agonizing over file system crashes, data corruption and disappearing disk LUNS. Positioned correctly, this team has the potential to change the face of scalable HPC file system deployments.

Trackbacks

  1. […] Lustre appliances, you can read their full release here or insideHPC’s exclusive interview here.AKPC_IDS += "7209,"; Posted in HPC Hardware, New Installations, Storage by John Leidel | 0 […]

Comments

  1. Doug Logan says

    An interesting approach with a strong value proposition. I couldn’t begin to enumerate the number of instances where my customers complained bitterly about the administration headaches of LUSTRE while at the same time acknowledging that there was no practical cost-effective alternative.