This column originally appeared on TechTarget's Expert Answer Center as a post in Robert Crawford's blog. Robert served as the on-demand expert on the Expert Answer Center for two weeks in January to February 2006, during which he was available
to quickly answer questions on CICS application performance and design, as well as to write daily blog entries. Keep an eye on the Expert Answer Center for topics that could help your IT shop.
Around the middle of each decade since the '80's we start hearing the rumors that the mainframe is dead. Sure, it still processes a lot of data and serves as the back end of some very large companies, but its days are limited. Any day now, CIOs around the world will wake up and chuck the monolithic dinosaurs for something more flexible and cheaper. I know I took the threat seriously enough 10 years ago to take a career detour into writing C++ code on Windows NT.
But, somehow, the big machines manage to survive and end up stronger. IBM has done so well redesigning the hardware that I can't remember the last time we had a processor go down (knock on wood). MVS and other subsystems -- such as CICS, DB2 and IMS -- are exploiting the parallel Sysplex to the benefit of availability and maintenance. And, just in case you need them, we have 64 bits.
There are other things about the mainframe you don't really appreciate until you work for a little while on another platform. After all, how does Unix security stack up against something like RACF? Updating the Windows operating system really makes one appreciate SMP/E, not to mention how difficult it is to safely and completely install or remove software on that platform. And what about these platforms that require upgrades or new boxes when they reach 50% capacity? This is a far cry from a mainframe that barely breaks a sweat when the CPU is in the low 90's.
When I first started in 1981, everyone in IT spoke pretty much the same language. That isn't true anymore, given the complexity of our systems and the choices of platforms. When I venture out of my department, I often find myself explaining to people how mainframes work as if it were some giant mystery. Unfortunately, I'm a small enough person that I've used this to my advantage. For instance, I heard someone talking about some Unix software designed to balance and manage workloads. "Oh, that sounds a lot like WLM, which was based on SRM about 20 years ago," I gleefully pointed out.
Still, as much as we love our behemoths, we have to recognize there are some things they can't do. I've heard statistics that indicate an Intel machine can easily run Java many times faster than a mainframe. Mainframes are also too big and ponderous to be used for spot solutions or small applications. And, to be entirely honest, there are just some things that are better left to smaller machines. Remember the cartoon of the mainframe mouse that was as big as a golf cart?
If mainframes are going to last much longer, some things need to change. I humbly submit my suggestions:
- College-level programs designed to train students on the mainframe. IBM already is pursuing this option.
- Mainframe software costs are outrageous and only get worse as vendors charge for machine or data center capacity instead of actual usage. Software vendors need to back off before they kill the goose.
- IBM needs to reduce the causes and effects of Sysplex "sympathy sickness." These are the cases where one sickly LPAR causes problems for everybody.
- Expand or eliminate limits on the I/O subsystems. Granted, mainframe I/O is very good. However, there are enterprises rubbing against the 64K device limit. Sometime in the future we may also want MVS to hook up to the SAN.
- IBM should keep up the good work started with parallel Sysplex. Someday we may be able to apply and install maintenance in the middle of the week instead of Sunday morning. At my age, I need my sleep.
This was first published in February 2006