Billed as HP’s official launch of the architecture, yesterday’s lengthy presentation by the company covered the same ground as an HP marketing event three months ago, which was also about impressive plans for products not due to ship until 2006 or 2007.
We did that event in Europe, and word about what we said then did not get out as well as we thought it would, said Michelle Weiss, marketing vice president for HP.
So lacking any short-term fixes to its problems – such as a way to prop up the sagging server sales that have dragged down its storage sales, or any new storage products to launch – HP blew its grid trumpet again.
It’s clear from our financial performance in HP’s fiscal third quarter that our execution recently has not kept up with our aspirations, Fiorina said.
But our aspirations haven’t changed. We are….driving our strategy and direction in a way that will provide revolutionary new capabilities for our customers – capabilities that no other storage vendor today can provide, she said.
Neither can HP deliver those capabilities today in mainstream products, although its storage grid concept has been praised by analysts such as Nancy Hurley at the Enterprise Strategy Group and Joe Zhou at DH Brown.
The reality is that HP has some innovative technology that truly will enable a flexible, adaptive grid in a way different to any other vendor, Hurley told ComputerWire this summer.
The concept – similar to one sometimes touted far less vigorously by IBM Corp – is to bring to storage the same scale-out architecture that already provides low-cost and flexible computing in farms of bladed servers.
Under HP’s plan, today’s individual storage arrays will be replaced by grids of hundreds of interlinked smart storage bricks, and the company says these grids will provide huge scalability and flexibility. Management should also be simplified because the grids will present themselves as single entities rather than as today’s collection of arrays each of which has to be managed individually.
Each brick will run the same operating system, but will be able to be loaded with different plug-in software on the fly to allow it to be allocated to different tasks, such as handling block-level data, pumping out NAS-style file data, or dealing with archiving and data retrieval applications.
Costs will be reduced in part by using industry-standard components – HP marketing speak that mostly means Intel processors, although the company stressed that it is not wedded to Intel’s chips. Existing storage arrays already use Intel chips as controllers, but Mike Feinberg, CTO of HP’s storage division said: There are a lot of assumptions made in current arrays, and a lot of coupling processors with special hardware. We can get the same reliability without that coupling.
HP was unable to say whether it has made any public demonstrations of such a grid in action. But there are two HP products that it claims already embody the grid architecture. These are the clustered file system that the company launched this summer, and the RISS reference data store that HP acquired when it bought start-up Persist Technologies late last year.
Like other clustered file systems, HP’s Lustre-based software allows multiple servers to share the same data. But rather than being a grid product, its most important characteristic is probably that it is a cluster file system that is both backed by an OEM and supports Linux.
The RISS box fits HP’s description of storage grids very well, because it is powered by multiple controllers, and its power can be boosted simply by adding more controllers. But the device does not yet work in federation with any other grids.
HP OEMs the high-end Lightning disk array from Hitachi Ltd, which with new virtualization software can now incorporate disk capacity from third-party arrays, such as HP’s EVA device. With an EVA attached to it, the new Lightning would be an example of grid computing, according to HP.
That’s a fully blown-out grid. It’s not scale-out, but it does give the notion of a single image, Feinberg said.
