IBM has been a supporter of the Globus Toolkit for almost 3 ½ years now. We did a bunch of cooperative development work with Argonne National Labs and USC ISI on the Globus 3.0 implementation, and the propagation of standards in the web services and Grid services arena (OGSI, WSRF, WS-Notification, etc.). We've supported the Globus Toolkit from multiple points on the compass. We contributed funding to the actual development of the toolkit. We have consistently contributed development resources and implementation resources to the toolkit. And since GT 2.0, we've taken the public open source and featured it in an IBM-specific implementation on IBM platforms. We had a GT 2.0 port of the toolkit to Linux, and we have an IBM Grid Toolbox version 3.0 that we make available on our platforms for our customers, free of charge, and they can also buy support for it.

And we've stated that we're working on an IBM Globus Toolbox release 4.0 - a similar implementation that will come out early in the fall. So we've been on the Globus Toolkit bandwagon for quite a long time. We've featured the Globus Toolkit in some of our customer engagements in both the public sector and the private sector, with companies that wanted to exploit Grid computing to support collaboration, reduce total cost of ownership or reduce time to results. IBM also has a role in building the community of application developers, implementers and end users. We're really strong proponents of the standards-based open source toolkit.

GCJ: Are most the implementations in the commercial sector still in financial?

The Globus Toolkit has been very well-received in public sector projects - universities, government projects and science initiatives. We're seeing interest from the commercial sector in the types of compute-intensive scenarios you might expect -- from petroleum exploration, to digital chemistry, to financial services, to certain types of predictive insurance applications. So we're starting to see the beginning of Grid uptake in the enterprise, and kind of the way you'd expect -- it's happening first in areas that touch most heavily on science and analysis. We're also starting to see people with more traditional sorts of high performance applications that are more in the mainstream starting to explore Grid.

GCJ: What kind of role do you see services playing as open source Grid takes off in the enterprise?

If you think about it, companies that are deploying any open source implementation as part of their day to day production infrastructure are concerned not so much that they have "one throat to choke," or one vendor to depend on - but that they have some vendor to depend on. If they have support issues, they know that they can get the support that they need in a timely fashion. If you go back and look at Linux - Linux really made inroads as a server operating platform once the hardware vendors and the Linux distro companies like Red Hat and SuSe Linux got involved in providing commercially tested and commercially integrated packages with appropriate levels of support. I think we're going to see the same phenomena in Grid. It's going to take a level of integration, testing, robustness and companies standing behind the Grid middleware infrastructure to really move Grid from the periphery of enterprise computing into the mainstream. If you integrate this stuff into your production environment, you want to be very certain that if you have a problem, you have somebody to go to.

I don't necessarily believe that an enterprise is going to look for a single vendor though. Lots of these enterprises today integrate middleware from lots of vendors. They buy their database from Oracle, and they buy their transaction computing and web applications server from IBM, and they do a certain amount of integration on their floor. In some cases they depend on outsourcing vendors like IBM and EDS and Accenture to be that "one throat to choke" - and then their outsourcing vendor buys support from the various middleware and software and hardware vendors that they have in their shop. But I think it's more that they fear "no throat to choke" in the open source space, than having expectations of one vendor providing all the support.

GCJ: Why is open source so hot in the enterprise right now?

Open source software that's properly supported by vendors is a very attractive proposition from a number of points of view. Open source / open standards software provides an interoperability environment in which enterprise customers can go and find the function they want from a variety of vendors and successfully integrate it and expect that it will work together. They have freedom of choice. The other thing about it - open source and open standard software is vetted in a way that proprietary solutions are not. There it is for everyone to see, understand, de-bug, shoot at (from a security/integrity point of view) - it's vetted in the open sunlight. It gives enterprise customers a feeling of safety because of that vetting process. That's been true over time with Linux systems. And that's going to become more true with open source Grid middleware over time, because I think the value proposition is moving up the value stack. Customers more and more are looking to proprietary vendors to provide higher value than the commoditized middleware layer. Nobody sells TCP/IP stacks anymore. They're a dime a dozen, and you can buy a whole TCP/IP stack on a 2-chip set, and they're embedded in the most mundane devices around. Over time, we're going to see that simplicity line, that commodity line move up the stack to a higher level of functionality. And I think ultimately some of the lowest level Grid middleware, and especially the web services middleware falls in that space - that is likely to be most advantageously supplied by open source software, because at the end of the day it's a commodity.

GCJ: Not surprisingly, the large systems vendors have dominated the Grid discussions to date. What's going to make some of the other major vendors -- say, application vendors and networking players -- get on board?

When we think about the vendors that are participating - there are some interesting companies not participating yet. Cisco, for example. I think you'll see that change. Not just Cisco - there are a number of companies that have made their name in other parts of the distributed system infrastructure that I think are going to step up their involvement in Grid activities. Take Microsoft. Microsoft has smart people that recognize that in some sense, Grid is synonymous with the evolution of scale-out distributed systems infrastructure. Microsoft isn't going to allow itself to get painted into a corner with the respect to the evolution of distributed systems. There are already implementations of some of the core standard web services functions on Microsoft .Net - there are implementations of WSRF and WS-Notification out of the University of Virginia.

One thing that's going to help get additional companies involved will be for the Grid industry to sort from its many standards bodies and communities down to the few. Today we have a lot of open source communities in the Grid space. We have the Condor and the PPS schedulers. We have Unicore. We have Globus. We have things going on in the UK e-science community. It's to the advantage of the community to sort down to fewer stacks. I see Globus as a rallying point for open source Grid middleware, and I think it's to everybody's advantage to converge on a common set of low-level and mid-level functional specifications and standards. That sort of takes time. The Globus Toolkit is maturing, going now into its fourth generation, and the functionality is beginning to be improved and refined and tested more robustly. I think it's a good choice for core technology, but you have to be careful here, because it remains to be seen how the overall community will vote for what the right convergence standard is. I think GGF and the open standards community plays a big role in that.

close window