ACADEMIA

ActiveScale 3.0 Brings 3X Higher Performance and Predictive Self-Management Capabilities to New Panasas ActiveStor Storage Clusters: Panasas, Inc., the leader in clustered storage solutions for the high-performance computing (HPC) market, today announced the third generation of the company's operating environment, ActiveScale 3.0. The ActiveScale 3.0 operating environment adds several new features to the established PanFS parallel file system that Panasas has deployed in top companies and scientific organizations around the world. The ActiveScale 3.0 enhancements provide large-scale reliability and manageability, including new predictive self- management capabilities and an increase in client performance by a factor of 3X to over 500MB/s per client, and up to 10GB/s in aggregate. Key enhancements to ActiveScale 3.0 include the new predictive self- management capabilities of ActiveScan and ActiveUpgrade. ActiveScan continuously sweeps the data and disk media in the Panasas StorageBlade for discrepancies and proactively corrects the media defects. ActiveUpgrade enables all software upgrades to be performed online while applications are running, eliminating user downtime for upgrades. Together, both new capabilities ensure maximum data accessibility for users. "LANL has deployed more than one petabyte of Panasas ActiveScale Storage, so we have been looking forward to the new capabilities of ActiveScale 3.0," said Gary Grider, group leader of Los Alamos' High Performance Computing Systems Integration Group. "We were particularly interested in the new higher single client bandwidth, RAID upgrades, parallel reconstruction, and background health monitoring of the storage. The higher single client bandwidth capability in ActiveScale 3.0 will help accommodate larger individual compute nodes within our terascale and upcoming Roadrunner petascale supercomputers. The 3.0 RAID features have shown very good performance with many extremely difficult parallel I/O workloads. Additionally, the aggressive way that ActiveScale 3.0 addresses reliability at very large scale will help us continue to provide very high availability to our many supercomputers that are sharing our common scalable global PanFS parallel file system. We have been testing the new ActiveScale operating environment and parallel file system to ensure a successful upgrade for our production file system deployments. The ActiveScale 3.0 system will be deployed on the Roadrunner system, one of the first attempts at a sustained petaflop supercomputer." In conjunction with the launch of its new ActiveScale 3.0 operating environment and new storage management applications, Panasas also announced today the ActiveStor 5000 and ActiveStor 3000 Storage Clusters which is based on ActiveScale 3.0. "Customers looking for scalable storage solutions to match their ultra high performance compute (HPC) Linux infrastructure needs require scalability in terms of performance, capacity, reliability and ease of management," said Greg Schulz, founder and senior analyst at the StorageIO Group. "ActiveScale 3.0 software enables customers to address a wider range of applications and their associated storage performance requirements." With the introduction of ActiveScale 3.0 operating environment, Panasas now offers a comprehensive portfolio of high performance storage solutions for HPC environments with the highest available bandwidth, unsurpassed scalability, complete data availability and simple manageability, which allows organizations to more fully maximize their return on investment and reduce overall costs. ActiveScale 3.0, the base operating environment for all ActiveStor Storage Clusters, includes a number of additional new predictive self-management features in addition to ActiveScan and ActiveUpgrade that further differentiate the Panasas storage solutions. Those features include: -- Predictive disk management: This feature anticipates hard disk drive problems with automated failure analysis. Upon prediction of a Panasas StorageBlade malfunction, all data is moved to other StorageBlades avoiding reconstruction. -- Parallel reconstruction: In the event of a StorageBlade malfunction all Panasas DirectorBlades in a storage system cooperate in parallel to significantly speed up the reconstruction rate of conventional RAID controllers, completing rebuild of a 500 GB blade in less than 90 minutes versus a typical rebuild of 24 hours. -- Real-time monitoring of client load generation: This capability automatically identifies performance bottlenecks among storage users and notifies administrators avoiding cluster downtime. The Panasas ActiveScale 3.0 operating environment includes NFS, CIFS and DirectFLOW protocols enabling the Panasas solution to work in heterogeneous HPC environments. The integrated software/hardware storage solution combines a next-generation parallel file system and an object-storage architecture that is perfectly tuned to deliver superior performance right out of the box. The DirectFLOW capability offers users a fully-parallel data path to allow high speed, direct communications between the Linux cluster and Panasas storage. This direct data path eliminates costly performance bottlenecks that can cause the cluster to idle as it waits for data. The fully distributed object-storage architecture gives the Panasas ActiveStor Storage Cluster the ability to scale in performance and capacity by distributing storage processing activity directly to the storage devices. Additionally, the Panasas ActiveStor Storage Cluster provides a single global namespace for very high scalability and near linear scaling in performance as more capacity is added. "As the HPC industry begins the transition from terascale to next- generation petascale computing, the ability to drive down the costs and complexities of upgrading to a new generation of computing becomes even more imperative," said Victor Perez, president and CEO of Panasas. "The ActiveScale 3.0 operating environment, in conjunction with a new line of Panasas ActiveStor Storage Clusters, helps commercial, academic and government HPC organizations bridge storage infrastructure requirements with ease." The ActiveScale Operating Environment will be released to new and existing customers in January 2007 and will be showcased at the upcoming SC06 Supercomputer Exhibit and Conference during the week of November 13, 2006.