|HOME||PUBLIC LIBRARY||ANOTHER PERSPECTIVE||INFOPERSPECTIVES||CONTACT|
With its announcement of z/VSE 4, which will become available in March, IBM has executed a hat trick. It has brought new technical life to an operating system that has not been part of the swift flow of mainstream computing. It has brought the usage-based pricing model used by z/OS mainframe shops to the VSE users who had previously been neglected. And it has set the stage for fresh hardware initiatives by speeding the exit of users with older machines.
Z/VSE 4.1, to use the full name of the product, is the first member of the venerable VSE family of operating systems to take some advantage of 64-bit z/Architecture. For now, VSE uses real, not virtual, 64-bit addressing and only does so for some systems functions. There is still no support for 64-bit virtual addressing within z/VSE, nor does the software support real or virtual 64-bit addressing for user applications. Nevertheless, the change in IBM's attitude, which had previously strictly confined the VSE environment to 31-bit computing, is the camel's nose slipping under the tent. Users will expect more and, inevitably, will get more, if IBM sees sufficient demand or if third parties offer ways for VSE products to reach out, possibly through a surrounding z/VM environment, and prosper as a result.
It's not yet clear how the cadre of ISVs will react, particularly the smaller developers. Some vendors of VSE packages, which range from large middleware subsystems down to small utility programs, may prefer to languish in a largely unchanging environment, while others will reach for every possible opportunity to proffer new and improved versions of their wares.
Some kinds of products are bound to grow, if IBM permits. Sorting software, for example, benefits from the potential of larger real memory, if for no other reason than it can use the memory to process larger blocks of data and reduce the I/O that slows the sorting process. Sorting is central to many business processes, so it's likely to figure prominently in any changes that catch on among progressive VSE users. DBMS processing, which, depending on one's point of view, complements, supplements, or competes with sorting, similarly can grow to take advantage of all the memory IBM allows users to allocate. Environments with many tasks and which require a lot of swapping to move between them would also find further movement towards true 64-bit computing an attractive possibility.
Still, many VSE applications are very economical users of memory, because they have had to be. A larger address space might make little difference to users with such programs. These applications depend not only on architectural changes for their capability, but also on compilers and run-time environments. If IBM doesn't roll out rich language support, future extensions of VSE into 64-bit territory will be moot.
Some of the other enhancements to VSE offered by Version 4 will also take a while to sink in. IBM's latest VSE provides some improvements in device support, but some of these improvements are also becoming available in VSE 3, and it's unclear how many VSE shops have taken advantage.
Similarly, IBM's improvements in security technology and its TCP/IP stack are not developments that users can quickly load and use. VSE shops will each weigh the business advantages against migration and development costs, and no two come to the same conclusion nor choose the same schedule for implementation.
Finally, IBM has highlighted some interoperability features of the VSE environment by carrying them forward to VSE 4, including them in usage-based pricing, and generally restating its willingness to accommodate shops that include VSE along with Linux and Windows environments.
So, the IBM message is quite clear: VSE is no longer dormant, and VSE users will not have access to a larger and much more lively playing field.
One incentive for VSE shops that are understandably reluctant to upgrade hardware and software unless there is a very strong budgetary case in favor of change is IBM's decision to extend usage-based pricing to VSE and about a dozen related software products, including language processors and interoperability software that links VSE applications to external databases.
The new Midrange Workload License Charges (MWLC) scheme for VSE users applies only to users with z9 hardware, and excludes those with the smallest z9 BC system, model A01, which runs at 26 to 28 MIPS, depending on whose MIPS estimate you use. MWLC requires metering software, and the richest usage measurement scheme IBM offers requires not only VSE but also a surrounding VM (release 5.2 or higher) environment. VSE under VM is quite popular, but not universal.
IBM hasn't yet provided detailed pricing for VSE and the middleware VSE products that are part of MWLC, but by the time VSE ships all the facts will be available to prospective users.
IBM claims that the benefits for many users are well worth investigation and, most likely, adoption of the usage-based scheme, because it ties software charges to the processing power actually used by workloads run under VSE, not to the whole platform on which the VSE system is housed. When a shop chooses pricing based on actual usage rather than system capacity, the rest of a mainframe can be used for other environments or for nothing at all.
Users who don't fill a machine are not punished for purchasing boxes that provide room for growth. This has proved to be attractive in the z/OS world, where IBM first developed its usage-based measurement and licensing plans, and it will undoubtedly draw interest from VSE shops, and user sites with a mix of VSE and other operating systems on their mainframes.
The third fork of the VSE announcement that will stir up interest in the VSE base is the way IBM has fenced off its z9 processors. Only z9 processors, and only those faster than the model A01, can take advantage of software pricing based on the working capacity of designation software partitions. Users with z890 (and z990) platforms can exploit VSE 4 technology but not its most flexible software licensing contracts. Users with z800 (and z900) hardware are additional restricted in other ways, as VSE will only run in what IBM calls basic mode. IBM has not yet provided the technical details of this last limitation, but it's not unreasonable to presume that the full description of VSE will show that its limited use of 64-bit processor technology will be confined to systems no more than one generation old.
The variations in VSE from generation to generation have been designed to allow users with older platforms to test the new operating system before committing to a forward migration and also to spur that movement to IBM's most recent processors. But they do more than encourage hardware upgrades. They also set the stage for future hardware announcements.
The z9 product line completes the IBM's consolidation of mainframe hardware into a single product line. There are some small differences between the z9 BC and EC families, but, aside from performance differences, there is really very little difference to applications developers. Code can be moved between z9 platforms with relative ease. By contrast, migration between low-end and high-end mainframe products in older generations, such as movement between a Multiprise and a 9672 or z box, can require considerable effort on the part of systems (and sometimes applications) programmers. While it has always been easy to understand why IBM might have chosen to manufacture more than one line of mainframes to meet the needs of users whose performance requirements cover a very wide span, it has been difficult for users to puzzle out some of the differences between contemporary IBM mainframe systems. These differences were eliminated when IBM introduced the 9672 but reborn as 9672 engines became more powerful and the costs of entry systems rose.
IBM offers better compatibility across its System/i and System/p product lines. With all its emphasis on virtualization and prowess in processor technology, IBM has not always delivered mainframes that present a unified platform to all users of any of its mainframe software environments. When Big Blue offers its next generation of flagship systems, which might well use engineering techniques that more closely resemble those used in Power based servers (and, for all we know, even use CPU chips from the Power family) to help deliver better value. It might turn out that, before long, IBM offers mainframes that live in an industry standard (and IBM standard) equipment rack, something the company has not done since the days of the 9221 product line.
— Hesh Wiener January 2007