Panasas, a network storage solution provider based in Fremont, Calif., is launching its first product this week, but already has some key customer wins in the bioinformatics community: The company claims Terry Gaasterland’s lab at Rockefeller University and Gene Myers’s group at the University of California, Berkeley among the early adopters of its technology. In addition, Panasus just signed an agreement with Los Alamos National Laboratory for up to 620 terabytes of storage and also has a system installed at Sandia National Laboratory.
How did the small startup capture the attention of the scientific computing elite? According to Panasus CEO Rod Schrock, the company offers a “complementary storage system” to the Linux cluster architecture that is quickly gaining ground in scientific and technical computing. Most storage solutions for cluster environments — even network attached storage — create a bottleneck between the compute nodes and the storage where data is “staged” and “destaged,” said Paul Gottsegen, vice president of marketing at Panasus.
Panasus uses an approach it calls object-based storage clustering that manages data in large virtual objects, and not as small blocks of bytes. Each object has metadata assigned to it that allows it to be dynamically distributed across blade-like “object storage devices” that sit in a 4U-high system shelf. The Linux application clients then use the company’s DirectFlow file system to directly interact with the data without going through a centralized filer, Gottsegen said.
The system provides up to 30 times the data throughput (at over 60 Gb/sec) and seven times the random I/O (at over 300,000 SFS ops/sec) compared to systems such as EMC’s Celerra NS 600 and Network appliance’s FAS 960, according to Panasas. The company said that its use of commodity hardware also keeps the cost of its system lower than that of its competitors.
The company’s product, which it calls the ActiveScale Storage Cluster, will be commercially available starting Oct. 20.