With the end of NEON’s zPrime, System z users lost an effective way to lower mainframe costs. And make no mistake about it; zPrime was effective in lowering costs. A data center manager in France told Dancing Dinosaur that zPrime saved his company almost 1 billion Euros each year.
There was no magic to how zPrime achieved these savings. Mainframe software licensing costs and various other charges are reduced when the processing is handled by a specialty engine (zIIP, zAAP, IFL) rather than by the general processor. The zPrime trick simply expanded the range of workloads that could be run on a specialty engine far beyond what IBM approved. No surprise that IBM shut it down.
But there is an immediate lesson here: to lower System z costs, shift as many sanctioned workloads as possible to specialty engines. The cost per MIPS on specialty engines is lower and software charges are lower. Call that lesson one.
Beyond that, there are a number of sanctioned ways to lower mainframe costs. Some amount to nothing more than taking advantage of IBM’s latest deals—these days IBM appears always ready to deal, especially when it comes to new workloads. In other cases, it is more like Back to the Future—reviving or updating best practices of the past, such as competitively pricing and negotiating functionality from different ISVs. DancingDinosaur previously addressed this here when looking at the software tactics of best-in-class mainframe shops. It turns out there is a lot of variation in ISV pricing, and many shops aren’t getting the best deals.
DancingDinosaur’s favorite deals come through the System z Solution Edition program. Solution Editions are deeply discounted bundled packages consisting of the appropriate hardware, software, middleware, and three years of maintenance for a particular workload. The Enterprise Linux Server solution edition package, now available for the z114, probably will be the cheapest way to get a mainframe for running a Linux consolidation strategy. Another Solution Edition package for the z114 will likely be SAP.
Another option for reducing mainframe costs is application performance management (APM). In the mainframe world, APM revolves around proven best practices to minimize CPU resource consumption, especially during peak times. It involves both rescheduling when applications run and optimizing the code to run more efficiently.
“APM starts with profiling and understanding the way your applications use mainframe resources as they run—especially CPU. It helps determine whether they really need all the resources they are using and with this information you can then make focused tuning efforts in specific areas of software code within the applications and especially the database calls, which tend to use a lot of resources. It can reduce the CPU requirements to run your applications by an enormous percentage,” explains Philip Mann, a principal consultant at Macro 4, an APM consulting firm and tool provider.
Using the Macro 4 approach and tools, British retailer Tesco reports reducing MIPS consumption 10-15% in one project, which allowed it to avoid purchasing extra CPU capacity. The systems in question involved ordering and distribution as part of the company’s retail supply chain process. With the Macro 4 tool Tesco identified opportunities where changes to databases, systems software, and applications could generate MIPS savings.
Another vendor quick to try to capitalize on the end of zPrime is DataKinetics, which provides tableBASE, a tool that speeds processing by moving frequently called data to in-memory tables, especially when running Java SQL queries against DB2. It converts these queries to DB2 stored procedures, which the company claims can reduce resource consumption by 90%. How much that nets out in the end as actual cost savings is unclear, but it certainly should boost DB2 performance.
OK, none of these techniques will likely save you $1 billion a year. But, when budgets are tight any savings will help.