Death of the shamplex?

Has the shamplex met its end? If you’re really using a Parallel Sysplex for what it was intended – distributing a single workload across two or more physical machines for load balancing and 24×7 availability, then you can skip the rest. But if like many of us, your mainframes are a Sysplex in name only – you’ve got operlog in the coupling facility so you qualify as a Sysplex, and maybe you’re sharing a single RACF database, then IBM’s recent announcement of Country Multiplex has implications for you.

Why do people run shamplex’s? Anytime you have more than one physical machine in a location, you need to have them in a Sysplex in order to get IBM’s discounted software pricing. (The origin of the Sysplex software discount is topic for another post). So why have more than one machine? In most cases, the answer is your ISV software costs.

A common use for shamplex was to create a software ghetto. Assume production needs 1200 MIPS to run, and your developers need another 200. You could run everything on a single 1400 MIPS machine, but you’d need to license all your developer tools for 1400 MIPS. So the developers get their own 200 MIP machine so the licensing costs for the tools is reasonable, and you run the machines in a shamplex.

Another common use comes from acquisitions and mergers. Both companies have mainframes and you’re merging the datacenters. It would be nice to run all the workloads on a single CPU, but the ISV software products don’t line up. So you wind up with two machines which you have to put in a shamplex in order to get your IBM software bill under control.

County Multiplex is a new pricing metric from IBM that promises to end the shamplex. As long as you’re on recent hardware (z196/Z114 or newer), you’ll be allowed to gain the benefits of the Sysplex pricing discount without having to physically couple machines that are located in the same country. Even CPU’s in different datacenters can aggregate their MSU’s for pricing purposes under County Multiplex. So the shamplex as a purely IBM software costing exercise is probably dead, and now the focus will be on how to control ISV software costs. In the long run, if the ISV’s stick to the model of “pay for all the MIPS on the box”, Country Multiplex may help IBM sell more System Z’s as users begin to isolate their problem software products to specific CPU’s.