IBM’s Elastic Storage Productizes Watson R&D

By Charles King, Pund-IT, Inc. May, 21, 2014

IBM’s announcement of its new codenamed “Elastic Storage” solution, part of its larger software-defined storage (SDS) portfolio, is dramatic by most any measure. According to the company, “Elastic Storage” performs unprecedentedly, scales infinitely and reduces storage costs by up to 90% via automatic tiering—moving data to the most economic storage hardware. In addition, like other products in IBM’s SDS portfolio, Elastic Storage can access/process any type of data (structured, semi-structured and unstructured) on any storage media (including tape, HDDs, SSDs and Flash) on any platform (including IBM’s competitors) without geographical limitations.

How does Elastic Storage achieve this? Via technologies originally designed to support the IBM “Watson” system that performed so memorably on the Jeopardy! TV game show. In essence, Elastic Storage leverages IBM’s Global Parallel File System (GPFS) solution’s storage access, management and governance tools, along with server-side flash (utilized as cache) and storage virtualization that allows multiple systems and applications to share common pools of storage. Elastic Storage also supports OpenStack cloud management software (Cinder and Swift) and other open APIs, including POSIX and Hadoop.

So the new technology can support a range of computational processes, including real time analytics on virtually any form of data, regardless of the underlying storage technologies, platforms and vendors. Practically speaking, this means that rather than choosing/moving data to individual systems for transaction processing, analytics and other efforts Elastic Storage can automatically and simultaneously support multiple workloads on the data residing in storage pools without requiring duplication or transference.

So how viable or useful is “Elastic Storage” likely to be for day to day business? That depends on several factors, including configuration, cost and availability. IBM’s announcement positions the new offering as ideal for data-intensive applications, such as seismic data processing, risk analysis, weather modeling and scientific research. In other words, cutting edge workloads where the need for raw performance often overrides price considerations. As a result, we expect interest in the new solution will mostly reside, at least initially, among companies and organizations already familiar with IBM’s Watson, GPFS and other technologies.

Those groups fall well within the company’s traditional enterprise and government research customer bases, and typically have access to the funding required for high performance technologies. But that situation could change notably after IBM begins delivering Elastic Storage via its SoftLayer cloud service infrastructure later in 2014. Depending on how those offerings are priced, designed and positioned, demand for Elastic Storage could eventually stretch beyond IBM’s initial target markets.

© 2014 Pund-IT, Inc. All rights reserved.