Issue link: http://osercommunicationsgroup.uberflip.com/i/265924
By Niso Levitas, Research and Development Manager, Aberdeen LLC Traditional hardware controller based arrays are sufficient for most applications. In time, RAID controllers embraced SSD caching to scale the performance a little higher, and there are all SSD NAS appliances for demanding applications. For supercomputing, or with a very large amount of data, the performance and capacity scalability may not be enough. Hardware based controllers are limited by their single purpose, small processors and 512MB or 1GB cache on the controller. Regardless of the drive type used, this is the bottleneck of the storage. Also, there is the data integrity issue. There are silent data corruption problems that may be an acceptable risk for most of the applications, but not for supercomputing. When you have a full server with two very powerful multi core processors and 256GB of memory that is mostly dedicated to drive your RAID arrays, you have your As datasets grow and both research teams and computation- al resources become geographically dispersed, tools that enable collaborative, long-distance data investigation play an increasingly critical role in scientific innovation. Open, scalable frameworks are necessary to readily facilitate the sharing, analysis and visualization of large data over the web and on mobile devices. Kitware is developing applications and libraries to meet these needs in several fields, bringing our robust visualization and analysis capa- bilities to developers and end-users in entirely new ways. Kitware is the primary developer of ParaViewWeb, a framework that provides embedded, interactive visualization capabilities for a variety of web environments. ParaViewWeb tackles the challenge of web visualization by leveraging the latest HTML 5.0 technologies to enable client-server communication with a remote visuali- zation node or cluster using a light-weight JavaScript API. With ParaViewWeb, web applications can easily embed interactive 3D visualization and data processing compo- nents, and researchers can readily collaborate through web-based scientific workflows. An interview with Phil Hughes, CEO, Clustered Systems Company, Inc. SCSD: Tell our readers about your company. What's your main line of business? PH: Clustered Systems integrates components, systems and software from leading suppliers with our patented two phase cooling infrastruc- ture to deliver the greenest high-performance compute systems. These whisper-quiet systems drive down infrastructure and operational costs. SCSD: What would you say makes your company unique? PH: Three things. First, our systems can go anywhere as no conditioned space is required. Second, we support much higher power densities thanks to our two-phase Continued on Page 17 Today's trend translates SMP as Shared Memory Processing rather than Symmetric Multi Processing, and the architecture for SMP is Numa – ccNuma, to be exact. Numa architecture systems for the enterprise mar- ket are being offered from several vendors, all based on Intel's upcom- ing "Ivy Bridge-EX" Xeon E7 processors. Numa architecture systems facilitate processing of the very large data sets that are common today (i.e., tens of terabytes), and offer superb new options for enterprises coping with large memory databases. Numascale is a Norwegian company that delivers a new take on this technology in its NumaConnect products: add-on cards built for standard volume servers which make it possible to build a large Numa machine at the price point of a cluster. The company's engineers have a long history of interconnect design, first designing the interconnect for the Convex Exemplar, a successful architecture that was continued by HP in its Superdome systems, and following that, the popular interconnect sold by Dolphin Interconnect. Continued on Page 17 SCALABLE LARGE CAPACITY STORAGE MEETS SUPERCOMPUTING KITWARE POWERS CUTTING-EDGE WEB-BASED VISUALIZATION CLUSTERED SYSTEMS OFFERS TURNKEY AND COMPONENT SOLUTIONS NUMASCALE OFFERS OPTIONS FOR BUILDING LARGE SYSTEMS Continued on Page 17 Continued on Page 17 O s e r C o m m u n i c a t i o n s G ro u p D e n ve r W e d n e s d a y, N o ve m b e r 2 0 , 2 0 1 3 AN INDEPENDENT PUBLICATION NOT AFFILIATED WITH SC