Sunday, 8 June 2008

Saving money with mainframes – revisited

I blogged about saving money with specialty engines on mainframes a couple of weeks ago, and I was taken to task over some of my conclusions – hence the need to revisit the subject.

I suggested that although specialty engines were themselves expensive, they would eventually save users’ money because workloads running on them don’t count against the MSU (Million Service Units) rating of the box. So software licences would be cheaper. I also said that a number of vendors were now supplying software that could utilize these specialty engines. And finally, I concluded that taking work off the main processor would save money because it would postpone the date when an upgrade would be necessary and also provide additional processing capacity.
Mark Fontecchio at Server Specs ( mentioned the success of IFL (Integrated Facility for Linux) as well as zIIP and zAAP. He doesn’t think that the savings in software licences are very much when compared to the actual cost of zIIP or zAAP specialty engine. He makes the point that the real saving comes from "being able to buy a six-figure specialty engine instead of a new seven- or eight-figure mainframe".

Marc Wambeke in his blog at Mainframe Watch Belgium ( says, "Saying that every workload which is redirected to zIIP or zAAP results in a reduction of your software cost is a bit too straight-forward. You still have to keep in mind that your monthly software cost (MLC) is based on a 4-hour rolling average. If you have your peaks at night running a heavy IMS batch workload, you won't see any direct savings on that side by adding a specialty engine." And it’s a good point to bear in mind.

Marc goes on to list the product he’s come across that make use of these specialty engines. If you know of any others, he’d probably appreciate an e-mail.

My other controversial method for saving money on a mainframe is to not run your software on it!! In these days of distributed processing, there are a number of products that take mainframe data and then perform most of the work on it on a different platform. Now, I know there are lots of arguments that the total cost of ownership of non-mainframe platforms is higher, so you should stay on the mainframe, but the reality is that data centres are full of boxes that aren’t mainframes. Just a couple of examples: Guardium for mainframes does most of its work on Linux boxes off the mainframe; and the CONNX suite can process mainframe data on distributed platforms including PCs. Just a thought!

And finally, Marc Wambeke pointed out to me that my blog on saving money on mainframes was also published at, but this version seemed to be attributed to vivin_bob. He may be a blog aggregator, but he ought to attribute other people’s work appropriately, don’t you think?

No comments: