Thursday, October 06, 2005

fyi GridWorld: Standards adoption key to grid computing growth


Pointer to article:,10801,105180,00.html?source=NLT_PM&nid=105180

Kobielus kommentary:
This article gets it plain wrong.

First off, there have been standards in the grid space for several years. The OGSA/OGSI standards that the article describes as “emerging” have in fact emerged a long time ago. In fact, OGSA/OGSI has been implemented widely, though not universally, throughout the grid market in various ways. In fact, OGSA/OGSI is well into its decline, have been superseded by a new set of WS-* grid standards, under the WS-Resource Framework (WSRF) umbrella.

Secondly, standards adoption isn’t what will drive the grid market. If standards were so important to grid adoption, vendors would have long since rushed to implement OGSA/OGSI and/or WSRF universally, to define industry-standard implementation profiles for those standards, and to hold interoperability bake-offs far and wide. If the latent demand for grid computing—as one of many high-performance computing architectures—were so strong, why isn’t Microsoft, for example, flooding industry airwaves with its grid go-to-market message? Does it even have a grid go-to-market message?

Thirdly, the so-called “dearth of [grid] business applications” is nonsense. Grid is just the next generation of massive parallel processing, clustering, and high-availability computing, and is serving the same core enterprise requirements. Grid’s aren’t an “architecture in search of applications”—they’re an architecture looking for applications that are ready to migrate from older approaches that address the same core requirements. And grids are making persistent inroads into the enterprise market, despite what this article claims. Just talk to any grid middleware vendor, and they’ll give you a lot of juicy enterprise case studies. (Grid middleware—“griddleware”?—pancakes, Canadian bacon---it’s morning, I’m hungry).

Increasing grid adoption depends on a long-running, inexorable architectural shift toward virtualization. Fundamentally, grid is a virtualization approach that abstracts applications from the distributed hardware substrate of processors, storage devices, and other devices across which the apps execute. It’s growing in importance and adoption alongside other virtualization approaches, such as service-oriented architecture (SOA). Actually, you can regard grid as an implementation of SOA, where the “service” is a pool of unseen but seemingly unlimited computing power. A distributed processing “blackbox” that encompasses an entire constellation of online resources. The time-honored time-sharing CPU-sponging Xanadu dream that refuses to die.

It will take time—perhaps another 10 years—before distributed computing environments shift decisively toward ubiquitous virtualization, SOA, and grid computing. But it will happen. The need for cross-platform architectural flexibility, scalability, performance, and availability demand it.

By the end of this decade, even Microsoft will realize their need for a strong grid message. Perhaps in “Black(box)comb.”