It is now crystal clear that IBM wants to be regarded as the prime guardian of Linux within the IT industry. Not that it plans to mould the Open Source Software into an IBM-proprietary product, of course; that would defeat the object. IBM sees its role in the Linux world as that of promoter, champion, benevolent uncle, and principal financial backer. Its reasons for this stance have been widely discussed - Linux is the best anti-Microsoft weapon around; it has vast support across the industry, and it offers IBM the chance to open up its proprietary systems in a way that would have seemed unthinkable a couple of years back.
Whether IBM will be proved right in backing Linux up to the hilt remains to be seen, but for large enterprises running a heterogeneous mixture of systems, the lure of the software is considerable. With the AS/400 version about to hit the streets, Linux is effectively the first cross-platform operating system that IBM has ever beeen able to offer. This has dramatic implications for application portability, of course - the Linux APIs and toolsets are identical on each platform, which means that in most cases applications can be moved from one system to another with nothing more than a re-compile. (Linux for S/390 is also ASCII based, which overcomes the problems with translating EBCDIC character sets). Linux also opens up great opportunities for companies to consolidate large numbers of, say, distributed web servers onto the centralised host, and to run a vast range of business applications that would be unavailable under OS/390 or OS/400.
IBM's recent marketing material leaves the reader in no doubt that the Open Source policy is driven from the very top of the organisation. The high-profile R&D investments in Linux and the powerful rhetoric surrounding an operating system that appears to cost nothing, and evokes all the altruistic harmony of San Francisco in the '60s; these are the signs of a marketing giant at work. IBM loves to play to the boardroom, and the Linux message is one that would gladden the heart of many an IT-weary chief financial officer.
The view from the bridge might be very rosy indeed, but down in the engine-room a degree of caution is needed. There are many critical decisions to be made about how and when Linux is implemented. Linux is an immature and untried product in the commercial business-critical world; Xephon's research suggests that only one in four data center managers have had direct involvement with the product.
Even the implementation choices need very careful consideration. On the mainframe, there are four options, native single image, LPAR, and VM - either the full VM/ESA, or the cut-down Virtual Image Facility. Each of these options raises a whole list of technical support issues, in terms of operational management, resource sharing, error recovery etc. VM/ESA is clearly the most functional implementation, and allows the consolidation of literally thousands of Linux images, but it is also the costliest route and won't appeal to shops that have already moved away from VM. Xephon's very recent research (stay tuned for more details!) suggests that many existing mainframe customers are going for the LPAR option initially, which makes a lot of sense as a 'toe in the water', but won't offer an economically attractive long-term option.
The other technical issue that users need to address is the balance between enterprise Linux and existing mainframe applications. Compared with the superlative workload management, data recovery, storage management, and performance strengths of the latest OS/390 and z/OS implementations, Linux is a mere pedestrian. Even compared with business-critical Aix and Solaris, it has a long way to go. We foresee a very long period of co-existence between OS/390 apps and Open Source. But Linux is relatively cheap, even compared with new usage based pricing schemes on the mainframe, and it carries a vast portfolio of leading-edge packaged applications in its wake. Established enterprises will treat Linux with understandable caution in the medium term, and will migrate function only when they can do so without compromising performance or manageability. But over the longer term, as the platform mature, who knows? We may see some real heavyweight OLTP and database tasks moving towards Open Source.
And this, for IBM, is a two-edged sword. The company argues that it can absorb the loss of OS revenue through the expansion of Linux middleware and support services. This might be true, if high-end Linux is taken up at a slow and measured pace. But, as the Chinese might say, we live in interesting times, and the recent economic downturn will make the bean-counters even more enthusiastic about potential cost-savings. More than ever before, IBM needs to work closely with enterprise users to establish what can and can't be done with Linux, and at what price. n
Mark Lillycrop is director of research at market analyst Xephon, and can be reached at firstname.lastname@example.org
This was first published in May 2001