Although a great deal of legacy angst is being expended on slow-developing Unix virtualization and workload management, equally important changes to Unix will come from simple availability defect analysis--with changes, for example, to memory redundancy mechanisms and PCI bus error improvement that minimize machine check interrupts and system termination. (In fact, system failures are being hunted down and eliminated, much as IBM did with its MVS operating system in the 1980s and 1990s.) This is not to say that Unix and z/OS will evolve to the same state. For example, catastrophic processor failures, which z/OS handles easily, are not on any Unix radar. Wisely, Unix vendors will not spend the considerable software dollars or significant hardware effort to implement, because hardware is becoming more reliable all the time. Unix vendors also will not provide the degree of virtualization found in z/OS.

[an error occurred while processing this directive]

In this case, both z/OS's hardware and software were built from the ground up, with virtualization in mind. For example, the ability to define parts of real and virtual processors to service a region or application did have "legacy" value, because operations groups tried to utilize expensive MIPS, driving even more expensive software. Unix MIPS are now considerably less expensive and, by 2007, will be 15x less expensive than z/OS. (Unix MIPS costs versus z/OS are 4:1; Unix performance is improving 35 percent per year and z/OS just 15 percent, resulting in a 15x ratio by 2007.) In fact, during this transition, legacy operators will have to adjust their "virtualization thinking," because they will be moving to significantly less expensive MIPS, where good-enough solutions (e.g., whole processor segments) will be reasonable for 90 percent of the services provided. Factoring in absolute cost and price/performance, good-enough Unix virtualization in 2007 will accelerate the z/OS slide in annual capacity growth rates to just 5 percent.

A few "by the ways"
Although virtualization and WLM have strong user interest, behind the scenes, Unix availability changes will be important for platform success. Error detection and logging, which are critical to system problem determination, will enter service in 2002 and, by 2005, 70 percent of high-end users will be exercising service-like processors for predictive failure analysis. Coupled with predictive analysis is graceful fault recovery (an emerging feature in 2002/03), and all vendors will provide a starter kit for isolating and deconfiguring components, such as processors, without a system reboot. However, to eliminate a significant number of failures induced by its "primitive" I/O subsystem, vendors will need an upgrade to improve error handling via PCI 2.0-like solutions (2002-04), and then a more independent I/O solution (e.g., InfiniBand) by 2005 if Unix is to be new mainframe. And although self-healing and self-management are key buzzwords, all vendors will be adding system monitoring points to automate availability and configuration data collection. By 2005, this will be a key platform differentiator.

Business impact
A well-defined/executed migration to maturing Unix platforms for high-end complex legacy systems can provide 20 to 40 percent savings in the 2005-07 time frame.

Bottom line
With strengthened workload management, dynamic partitioning, and improved availability, high-end Unix vendors have amassed critical hardware and software strength as well as direction for a clear shot at providing good-enough mainframe service by 2007.

Unix on Schedule to Look, Feel, and Act Almost Like a Mainframe by 2007
First published October 7, 2002
By Rich Evans

Are you considering migrating your company's high-end legacy systems to Unix? TalkBack below or e-mail us with your thoughts.

Previous page |   1 2

Newsletters
	Tech Update Today eBusiness Update	Tech Update Weekly		All newsletters FAQ Manage my newsletters

[an error occurred while processing this directive]

[an error occurred while processing this directive]