History 2004: Grid Storage or Storage Grid?

Grid storage or storage grid? Neither of these choices has emerged as a clear favorite, but we’d opt for the first if pressed, which recalls a known technology with which it shares several similarities, grid computing {no one calls it a computing grid). Others prefer the term ‘storage cluster.’

In the history of disk subsystems, grid storage may ultimately prove as important as RAID technology, so widespread today. And yet, if the invention of RAID can be exactly situated, in 1987 by students at UC Berkeley, the origin of the grid storage concept is much vaguer.

We know that IBM worked on the idea for several years. NetApp and HP are among the leading storage companies that have adopted the name, although with a slightly different definition. As has EMC, although without saying so.

In grid computing, the resources of various remote computers are utilized in parallel in order to distribute the processing of a single application. This requires software that divides and farms out pieces of a program to several computers. In grid storage, the data are stored in independent storage modules, not necessarily all identical, linked together by a network (typically Ethernet) to form a highly available system with unlimited and dynamic scalability.

Here again, a software virtualization layer, actually a common file system, on top of the system hardware, manages the overall grid – without the need of large central switch or hub. The storage modules may be NAS or removable and interchangeable storage bricks.

Redundancy is possible within each module or between modules. Start-up venture ExaGrid Systems prefers to call it RAIN (Redundancy Array of Inexpensive Nodes) rather than grid storage.

Storage on a grid could work with a number of applications that require enormous storage capacity, these days mainly for secondary storage, such as D2D backup, CAS, archiving, etc. Nothing prevents its use for primary storage, however. The technology of relatively cheap storage modules connected together to build a high-availability storage subsystem, thereby avoiding monolithic cabinets with expensive controllers, has every chance of making a decent splash.

It has already been embraced in EMC’s highly successful Centera, on NetApp’s NearStore backup/archiving appliance as well IBM’s forthcoming (although not tomorrow) IceCube, for which the term “Collective Intelligent Bricks” is preferred. At Big Blue, each brick is a complete PC in the form of a cube, connectable to another on all sides by proprietary wireless couplers rather than Ethernet.

Other companies besides HP involved in hard- or software grid storage include: 3PARdata, Bycast, Exanet, ExaGrid, Isilon Systems, NetApp, Panasas, Permabit, probably Pillar Data Systems, PolyServe, Rainfinity and Xiotech.

This article is an abstract of news published on issue 197 on June 2004 from the former paper version of Computer Data Storage Newsletter.