Meta trend
By 2006, OS/390-z/OS's onerous software costs (two to five times those of high-end Unix and about 10 times those of Windows 2000) and 15 to 20 percent annual hardware price/performance improvement will slow the annual net capacity growth of S/390-zSeries to less than 15 percent (versus Windows 2000 at more than 60 percent and Unix at 50 percent). By 2011, the capacity composition of enterprise data centers will approach 48 percent Windows 2000, 42 percent Unix/Linux, and 10 percent OS/390-z/OS (legacy).

[an error occurred while processing this directive]

Our research indicates that, although only about 10 to 15 percent of mainframe users have migrated to IBM's 17-month-old flagship z/OS, most of those did so to exploit its exclusive 64-bit mode on zSeries hardware. Our research also indicates users were intimidated by the perceived high risk of one-shot z/OS migration to 64-bit mode. The inability to back out to 31-bit mode, if required, is an issue addressed by IBM's new Bimodal Migration Accommodation. We believe that, during the next 18-24 months, the principal z/OS migration driver will be users' business case for 64-bit mode. Our research indicates that the key incentives for users to make the jump to 64-bit mode center on relief for database memory constraints, improved intra/intercommunications (e.g., HiperSockets), and partition and I/O management (Intelligent Resource Director--IRD). Moreover, z/OS R1.4 is reaching reasonable maturity/stability (most of the many annoying glitches characteristic of z/OS's first three releases have been fixed in R1.4), and will be the last z/OS release to which OS/390 R2.10 users can migrate directly. Users should understand that full 64-bit mode is a key prerequisite for IBM's much-touted Workload License Charge software pricing, which gives users the ability to more closely tailor software costs to usage.

Mainframe market futures: the few and the huge
Meta Group projects that, by 2007, the current mainframe market will have winnowed down to a relatively few very large Global 2000 users, where 10,000-plus MIPS centers driven by 500 MIPS uniprocessors will be the norm. We believe the mainframe market's current low to midrange (less than 1,000 MIPS) will inevitably be driven toward one of the following three alternatives:

• Unix/Wintel-based systems (60 percent): The majority of low-end users (especially those less than 500 MIPS) will eventually succumb to the "push" of unaffordable mainframe cost of ownership and the "pull" of more competitive, cost-effective, and adequately robust Unix- and Wintel-based platforms.
• Mainframe outsourcing (30 percent): A significant minority of current low-end mainframe users will find outsourcing their mainframe operations to lean, efficient, large-scale outsourcers (e.g., EDS, CSC, and IBM) a cost-effective alternative. Such outsourcing transactions will be either a long-term commitment (outsourcing a stable, non-strategic platform) or an interim step in a proactive long-term plan to migrate off the mainframe.
• Organic growth (10 percent): A small minority of current low-end mainframe users will grow out of the unaffordable low end, reaching not unreasonable economies of scale in the 2,000 to 5,000 MIPS range.