By Robert Woolery, VP Corporate Development, DataDirect Networks -- The ability to accelerate application performance when scaling is one of the keys to push supercomputing to higher and higher levels. This capability to build larger, scalable disk-storage facilities to handle the output from these large multi-teraflop scientific computers without degradation in application performance requires project managers to find enabling network technology that can meet their performance demands, while bringing less complex and more easily managed infrastructures. Many supercomputing facilities are determining that storage network appliances represent a new class of infrastructure that can deliver this capability. Seen by a growing number of project managers, administrators and integrators as a truly enabling network storage infrastructure device tailored to address the specific applications needs within supercomputing, storage network appliances accelerate applications such as visual simulation, modeling and rendering, storage consolidation, rich media management, and storage resource management. But as important, storage network appliances reduce the cost of acquisition, implementation and ongoing management with its simplification of complex infrastructures. Enabling a new and innovative architecture, the storage network appliance permit a much more efficient, scalable, and higher-performance Storage Area Network (SAN). In fact, appliances such as the Silicon Storage Appliance can allow supercomputing centers to accelerate their applications, providing data at least three times faster than existing SAN technology and ten times faster than Network Attached Storage (NAS) technology. The architectural simplicity and performance capabilities of appliances can provide unmatched benefits not possible in any competing amalgamation of discrete storage devices. For example, this simplicity with performance enhances a switched fabric by reducing performance impediments due to switching latency and contention, while mitigating hard costs incurred by cascaded fabrics. As many HPC professionals are learning today, the storage network appliance is the answer to the hard problems associated with building a bulletproof, cost-effective, scalable storage networks. Silicon Storage Appliances Just as the evolution of the computer networking market necessitated purpose-built routers and switches, the evolution of the storage networking market is necessitating a new integrated platform in order to meet performance, high availability, scalability and ease of management requirements of supercomputing data centers. The Silicon Storage Appliance, one of the new breed of storage network appliances, is a plug-and-play infrastructure device that accelerates application performance, simplifies information management, and reduces TCO that illustrates the blending of benefits that deliver results. The appliance integrates seamlessly with legacy storage devices while providing a path to cost-effectively scale storage capacity and virtualize storage pools. The appliance platform is available in a variety of configurations to serve departmental, enterprise data center and carrier-class environments and the platform can scale to accommodate higher throughput and storage capacity simply by clustering multiple units. The appliance platform is extensible beyond local storage networks and can play an enabling role in metropolitan and wide area network applications as well. As a new and different architecture, the Silicon Storage Appliance enables much more efficient, more scalable, and higher-performance SAN. The appliance, in general, can be used as a “simple” storage consolidation device, or when using a SAN file system as a file-sharing device. The unique virtualization, parallelism, and LUN performance characteristics of the device holds special advantages for file-sharing configurations. The unprecedented parallel nature of the Silicon Storage Appliance also allows for exceptional High-Availability and Load-Balancing features when using “multi-pathing” capable software such as Veritas Dynamic Multi-Pathing (DMP) or driver-level implementations such as those available from QLogic and Troika Networks. Storage Networking Appliances in the Supercomputing Centers The compelling nature of the storage network appliance is being demonstrated everyday, worldwide, by HPC professionals using them to solve a range of supercomputing problems from storage consolidation and backup acceleration to simplified storage management and increased application productivity. Below are a few examples of how different appliances are solving a range of HPC storage problems. Application Acceleration Sandia National Laboratories, part of the Accelerated Strategic Computing Initiative (ASCI) Tri-Lab program, is a research and development facility with ongoing projects in national security, energy and environmental technologies. Sandia is developing a large, scalable disk-storage facility to handle the output from large multi-teraflop scientific computers. To meet increasing data and high availability demands needed in their projects, Sandia turned to storage network appliances to accelerate visualization and rendering work that results from the complex applications that simulate, analyze and predict performance of nuclear weapons. The Silicon Storage Appliance provided the answer for simulation, analyzes and prediction acceleration, while delivering a highly scalable, simple-to-deploy and easy-to-manage storage network. One of the top line goals, according to Sandia’s project manager Milton Clauser, was the ability to easily handle very large datasets at very high I/O rates for consolidated storage pools – and this was a key consideration in the selection of Silicon Storage Appliances. “After considerable testing on a range of storage systems,” Clauser said recently, “Silicon Storage Appliances brought performance and management benefits that were very compelling.” Sandia ultimately selected appliances that each supplies an aggregate bandwidth of 800 megabytes per second and support up to 512 servers and 180 Terabytes of storage to bring ease of management, performance and scalability benefits. Silicon Storage Appliances have allowed Sandia to consolidate storage and easily deploy massive clusters to achieve substantial performance gains over earlier, more expensive technologies. When final results were in, Sandia acquired six Silicon Storage Appliances with 33 Terabytes for production use with existing computer systems and for development of new data services for future, even larger computer systems. Storage Consolidation NASA Goddard is a research and development facility with ongoing projects in satellite imaging and data analysis. NASA has created a 110+ TB storage network environment comprised of Linux, SGI Origin and Sun servers, with a series of Silicon Storage Appliances at the core of the network. With minimal administrator oversight, NASA is able to manage 110+ TB of storage and the storage network bringing substantial Total Cost of Ownership (TCO) and productivity gains for the environment. NASA selected Silicon Storage Appliances because they allow them to consolidate storage and easily deploy massive clusters to achieve substantial performance gains over earlier, more expensive technologies. Ultimately, NASA looks to converge SAN and NAS infrastructures through the use of appliances to bring data access to tens of users throughout the facility. Simplified Storage Management Lawrence Livermore National Laboratories (LLNL) also selected Silicon Storage Appliances to use in critical national security work. Livermore purchased a total of 12 storage network appliances and 91 Terabytes of storage to use in conjunction with the Accelerated Strategic Computing Initiative at their national lab. Livermore employed Silicon Storage Appliances because they exceed LLNL’s performance expectations with its scalability and management ease. Additionally, an LLNL executive praised the ease of configuration and maintenance, and predicted that the appliance “should reduce administration time and provide operational savings" to the national lab. Substantial scalability benefits, combined with the simplicity of deployment and management make the Silicon Storage Appliances a very compelling proposition to government, scientific and commercial computing environments. In these economic times, organizations are even more keenly attuned to technology – like Silicon Storage Appliances – that can help companies work faster, smarter and easier, reducing TCO and increasing return on investment. Enabling Storage Networks through Simplicity and Performance As a new class of storage infrastructure device, the Silicon Storage Appliance has capabilities that are simply not available in any other storage architecture. These capabilities result in a better cost-of-ownership, a more easily managed, less complicated, higher-performing and more robust storage network. The Silicon Storage Appliance is equally adept at solving problems of storage consolidation, storage management, file-sharing, accelerated backup and restore, highly available fault tolerant solutions and increased application productivity. The architectural simplicity and performance capabilities of appliances provide unmatched benefits not possible in any competing amalgamation of discrete storage devices. This simplicity and performance enhances a switched fabric by reducing performance impediments due to switching latency and contention, while mitigating hard costs incurred by cascaded fabrics. As many HPC IT professionals are learning today, the Silicon Storage Appliance is the answer to the hard problems associated with building a bulletproof, cost-effective, scalable storage network. ---------- Robert Woolery is vice president of corporate development for DataDirect Networks and can be reached through e-mail at rwoolery@datadirectnet.com. More information on DataDirect Networks and storage network appliances can be found at www.datadirectnet.com