Scalable Informatics, today, announced the release of their new siCluster storage cluster platform. The new storage platform is aimed at providing users with the ability to reliably scale their storage capacity and performance. The basis for the new platform is Scalable Informatics’ JackRabbit.
By leveraging JackRabbit high performance, tightly-coupled storage or Delta-V value-priced storage, siCluster enables you to scale uncompromising per-node performance and massive per-node capacity to whatever level you require, whether it’s two terabytes or hundreds of petabytes of raw storage,” said Dr. Joseph Landman, CEO and founder of Scalable Informatics.
Scalable provides configurations that include high performance network fabrics such as Infiniband and 10GbE. It also incorporates GlusterFS in the standard configuration.
The bandwidth wall, common on many other storage designs, is absent in siCluster,” noted Dr. Landman. “An 8-node JackRabbit JR5-based siCluster has been measured providing 16GB/s read and 12GB/s write across 32 nodes for 32 simultaneous IOzone processes. This performance is becoming increasingly needed in today’s research, and business computing environments.”
Alongside the platform release, Scalable has also announced their first major delivery of siCluster. The Minnesota Supercomputing Institute is the proud new owner of a GlusterFS-based siCluster. According to the release, the install was designed to provide a minimum of 8GB/s of sustained write performance.
The Scalable Informatics storage solution provides MSI with a high-performing, scalable storage system for our newest supercomputer,” said Brian D. Ropers-Huilman, director of systems administration and technical operations at MSI. “Scalable’s design and choice of the Gluster file system will give MSI the bandwidth to allow our researchers to utilize Itasca with 8,664 cores at full scale. Scalable’s system offered the best price to performance to capacity ratio to MSI.”
For more info on siCluster, check out the Scalable Informatics page here.
