In-Depth

IBM Touts Linux-on-zSeries Grids

Grids demonstrate mainframe value and versatility

• By Stephen Swoyer
• 04/17/2003

Big Blue has had some success selling mainframe hardware in support of grid computing applications—most notably, to the University of Florida—and, analysts say, is apparently positioning z800 and z900 mainframe hardware for use in specialized grid computing environments.

In this respect, IBM can point to a range of grid computing solutions that are now available for Linux-on-zSeries. SuSE, for example, has for some time offered a Linux distribution designed to run on top of Big Iron, first for IBM’s S/390 mainframes and now for its zSeries systems as well.

SuSE also announced that version 2.2 of the Open Source Globus Toolkit is available for Linux-on-zSeries as part of its Enterprise Server 8. The Globus Toolkit provides a set of tools or components that can be used either together or individually to build grid applications or programming tools.

While SuSE has been a Linux-on-zSeries player almost since the beginning, grid computing specialists Platform Computing and DataSynapse have traditionally marketed software designed to support grid computing applications on non-mainframe platforms.

Platform Computing, for example, develops Platform LSF, Platform Job Scheduler, and Platform MultiCluster, tools that support grid computing, workload balancing, and job scheduling. DataSynapse, on the other hand, markets LiveCluster 3G, a grid computing and job scheduling product.

Both companies entered into Master Relationship agreements with IBM in February to jointly develop grid computing solutions. (See (http://www.esj.com/news/article.asp?EditorialsID=399 for details.)

The announcements from SuSE, Platform Computing and DataSynapse came only weeks after Big Blue celebrated the sale of its 1,000 z800 “baby” mainframe—in slightly less than a year’s time—to the University of Florida to support highly specialized grid research. (See (http://www.esj.com/news/article.asp?EditorialsID=448.) Unlike most grid computing implementations, which exploit discrete systems or system partitions, the University of Florida’s proposed grid project emphasizes the virtualization of technology at the machine, network, data, and application levels. The idea, researchers say, is to facilitate the creation of dynamic virtual information grids on a per-user or per-application basis.

At the time of the announcement, zSeries GM Erich Clementi said the University of Florida selected the baby zSeries mainframe because of its sophisticated virtualization capabilities.

In light of these developments, some industry watchers suggest that IBM may begin pushing its zSeries mainframes for use in esoteric grid computing applications, such as in the hyper-virtualized environment the University of Florida will attempt to create.

At the same time, they stress, Big Blue probably doesn’t expect customer enthusiasm for Linux-on-zSeries application grids will result in a large number of new customer wins. Writing in a recent Gartner Inc. news bulletin, for example, analysts Carl Claunch and Mike Chuba were quick to dismiss the possibility that a significant number of enterprises would likely purchase “new zSeries mainframes to serve as the basis for grid computing.”

Both analysts also speculated that many mainframe shops have too few idle cycles to support grid applications. Nevertheless, they concluded, “some unique situations may arise”—such as grid applications that require a high-degree of virtualization, or distributed tasks that demand high availability and reconfigurability—“where zSeries mainframe[s] may have some direct benefits.”

Gordon Haff, a senior analyst with consultancy Illuminata Inc., agrees. “Obviously, this isn’t something that IBM envisions as having mass market appeal. I don’t think IBM even thinks that it will result in very many, if any, new customers wins.”

Instead, Haff and other analysts say, IBM’s Linux-on-zSeries grid computing push should again demonstrate to mainframe customers the virtuosity of their Big Iron hardware and IBM’s role in enabling it. The upshot, suggests Haff, is that existing mainframe customers will feel more confident about the viability of their investments—and will perhaps plan new mainframe purchases as well. “IBM may sell some upgrades, they may even sell some additional mainframe systems, but it’s primarily about sort of effectively reusing assets that are already in place, which is not a bad thing, of course, particularly in today’s climate.”

For example, because Platform Computing and DataSynapse provide grid computing products for mainframe and non-mainframe platforms alike, some organizations could opt to harness even their few spare mainframe MIPs in support of enterprise-wide grid computing efforts. This is possible because software from both vendors is able to schedule jobs in mainframe, Unix, Linux and Windows environments. Suggests Haff: “Certainly there is the potential for some very interesting applications there, which can … allow people to very effectively involve their mainframe assets. There’s the possibility that this could even result in new sales, as [IT decision-makers] realize how versatile these systems really are.”

In this regard, Haff specifically cites IBM’s next-generation G8 mainframe CMOS, which is expected to officially debut in IBM’s new “T-Rex” mainframe systems sometime late in Q2. “There’s always good reason for staying with a platform that you know and that you’re familiar with, especially when IBM continues to demonstrate the value in doing so.”

For their part, Gartner analysts Claunch and Chuba sound a similar note, noting that IBM’s recent moves add “yet one more justification for enterprises to retain or upgrade their zSeries installations.”