It's becoming clearer every day (unless Sun or Microsoft have something up their sleeves, and that looks unlikely) that there is now sufficient momentum behind the Globus Project and the Open Grid Services Architecture (OGSA) to drive them forward as the standard platform for grid computing – at least the handshake communications component. It appears likely this will help focus grid computing away from the discrete resources themselves and toward a new class of services in a 'post-Internet' computing paradigm.
By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
IBM: IBM's involvement in OGSA, Globus, Global Grid Forum and its general evangelizing activity has been instrumental in bringing grid computing out of the science lab and on to the mainstream IT agenda.
For IBM, grid is the next big thing. CEO Sam Palmisano identified grids as a key underpinning for its e-business-on-demand initiative, and the IBM divisions are stepping up to the plate with supporting announcements. But for IBM, grid computing is predominantly an IBM Global Services opportunity. Distributed, shared computing means changing the way companies implement and use IT, changing business processes and reorganizing resources. This will require significant professional services and consulting, quite apart from any technology. Indeed, utility computing will depend on customers' business models – which is why IBM is developing utility computing infrastructure and why its sales force has been energized with grid messages.
IBM grid computing general manager, Tom Hawk, sounded one word of warning – or at least a reality check – at last week's GlobusWorld conference, however. He said that while grid could ultimately deliver some profound changes in the way companies are able to do business, grids aren't the only solution out there. If, for whatever reason, grids can't step up, there are plenty of other competing technologies to fit the bill.
IBM sees 2003 as the year that will begin OGSA and Web services integration, point applications and commercial pilots. In 2004 there will be OGSA products and other utilities that ride on top – billing, metering, accounting and provenance. OGSA will sit in a container in IBM's WebSphere app server, and a version of the Globus 3.0 toolkit for WebSphere is under way. Web services and OGSA utilities – security, workflow, database, file system, directory and messaging – reside underneath, attached to OGSA-enabled servers, storage and networks. Five years out it expects compute, data and storage grids to be bound together with intelligent caching glue, although it has not been very specific about what this is. It says it will target global, enterprise, R&D, business analytics, engineering and design grid opportunities. Meanwhile, CPU cycle scraping doesn't have much of an enterprise grid future, IBM reckons.
With its supercomputing-on-demand effort, IBM clearly has per-GFLOP, per-hour metering charges in its crosshairs.
Sun: If IBM can claim to be leading the way conceptually, Sun is claiming de facto market leadership, by dint of the 7,000 grids it says have been deployed on GridEngine. The N1 datacenter OS is the vehicle for all Sun's grid plans.
Like IBM, Sun expects grids to eventually blend into the ordinary IT fabric, following the trajectory of other HPC technologies, such as RISC, SMP, parallel computing and CC-NUMA. The company believes there are a number of different levels at which grid technologies can participate. These include networked computing, which is entering a new phase where – whether called N1 or autonomic computing or something else – grids play a part; utility computing, either as a model for delivering resources within an organization or a way of IT suppliers selling computers/computing; partner grids, for sharing resources between trusted partners; and global grids.
In five years it is likely that networked and utility computing will be well established, and that partner grids will have some take-up – likely among academia (the obvious one) and ASPs. However, security and politics will nix any likelihood of full-blown global grids.
Sun sees standards (OGSA) and a Globus Toolkit 5 timeframe release as the right infrastructure to support the utility model. In an ideal world, the company would like to see a future GT release rewritten in Sun technologies – Java Jini and JXTA – with an infinite number of products and services on top.
Sun hasn't been able to marshal its grid activities with the same effectiveness as IBM or HP, nor has it demonstrated the same CEO-level commitment. It's surprising given Sun's need for ammunition in the wider war of perception it is fighting against Linux, Microsoft, IBM and proprietary technologies, as well as given Wall Street's clamor for Sun to return to selling innovation. Sun has said its chief competitive concern is that Microsoft will create a non-standard grid software that will throw the sector into confusion.
Everyone we've spoken with in the community says Sun's use of the term 'cluster grid' to describe a simple cluster (as opposed to a cluster of clusters) hasn't been very helpful, and indeed this proprietary view has confused the market.
Microsoft: Although a partner and sponsor of OGSA and Globus, Microsoft hasn't had much to say about its grid plans. Indeed, until it completes its Web services push, it's unlikely that we will hear too much. The company is not committing to use GT3.0 or OGSA, despite being a Globus sponsor, because it doesn't yet see any volume opportunity. Can grid computing ever dovetail with Microsoft's goal of selling more desktops and servers?
Microsoft's interest in grid computing right now is focused on data grids and providing access to SQL Server resources. With this in mind, we think the activities of IBM and Oracle in the data grid space will ultimately help spur Microsoft into action.
From a technical perspective, Microsoft's plan is to get all its products running on .NET within five years. A component layer on top will provide integration for .NET apps and high-level access to resources. Part of this is its services-oriented Business Framework, which will provide things like transaction and workflow utilities – things that sound very grid-like.
Microsoft believes utility models are achievable within the enterprise and to a limited extent outside of firewalls, but it doesn't believe global grids are, due to security and political concerns.
The451 assessment: There are challenges and opportunities ahead in the areas of security, provenance and the addition of utilities to take advantage of the handshaking. But the momentum behind the Globus toolkit and OGSA should help focus grid computing away from the discrete resources themselves and towards a new class of services in a 'post-Internet' computing paradigm.
The451 is an analyst firm that provides timely, detailed and independent analysis of news in technology, communications and media. To evaluate the service, click here.
CLICK for our Featured Topic on The451