Mainframe Platform Economics

Happy 2014 to all my readers, I hope you had a merry holiday with your families and friends and that you have a happy and prosperous 2014.

So with that out of the way, I would like to cover the thorny subject of Mainframe Platform Economics.  Despite your obvious audible groan as you read that last sentence, I hope to make it as light hearted as possible and poke some fun along the way...

So the Mainframe is obviously expensive and everybody knows their are better platforms to run applications on in 2014, heck the mainframe is 50-years old in 2014 there is no way it can compete with modern technology from a TCO perspective...

No, I have not been abducted by aliens, I just thought I would open up with the 'accepted' wisdom that us Mainframers hear pretty much every day from our IT brethren who look after distributed platforms. We have all heard these words or something similar when the subject of platform selection for any new application is being discussed.

Let me put in context some of the things such simplistic statements over look or more sinisterly ignore on purpose:

Total Cost of Acquisition Vs Total Cost of Ownership
Since the financial meltdown of 2008 the cost of running businesses has come more strongly into focus, the role of the Accountant and ultimately the CFO has gained a level of involvement in investment decisions previously unseen.  Given this rise to the forefront of IT decision making, of cost, then it is not surprising that investment decisions need to be made soundly and with foresight to how they affect the cost base of the business over the medium to long term.  I think that most readers would subscribe to the view that financial prudence is good practice, unless of course you are Larry Ellison (bugger that New Years Resolution didn't last long).  So why given this background do I see an increasing trend to focus on upfront costs or Total Cost of Acquisition?  This is short sightedness or delusion depending on how much credit you want to give the person proposing this approach.  Looking at how much a server costs is obviously important and should be ignored, but neither should the fact that it has little impact on the running costs over 5-years.  Indulge me if you please, would you buy a V12 Jaguar XJS that was 15-years old for £3000 or a £4000 Ford Fiesta for your 18-year olds first car after they pass their driving test.  Now if you are part of the Jaguar Owners Club and have landed here by the vagaries of Google I apologise profusely, but if you are not then you will get the point.  We all know the Jag would be thirsty, expensive to insure and cost a lot more to service...  You get my point...

Platform Selection - Security, Availability and Scalability
Again we in IT all know that a platform that is insecure, always down and has limited scalability should never be purchased... So why in so many business cases this author has seen over the last 19-years are these critical factors at worst not covered or at best glossed over.  When a business case looks solely at the year one costs then be wary.  We all know that applications grow and increasingly issues such as Security need to be taken into account up front.  Why then are these two areas not given the focus they so richly deserve?  Could it be that if we include security and scalability provision the TCA would suffer and the CFO would never sign off our business case and they don't know about IT so what the heck... Availability is another ball game altogether, take a look at this very interesting Wikipedia link:

http://en.wikipedia.org/wiki/High_availability

Once you have read this, is Amazon's new Cloud service that offers 99.5 availability still such a good decision.  What are you going to do for instance for the 50.4 minutes every week when your mission critical platform is down, put the kettle on? play Sudoku? apply online for a new job?  It really is not that difficult to put a cost on availability. One simple way would be to note the revenue/profit generated by an application down and then calculate how many hours in a year (365*24) and then work out what an hour of downtime costs.  I know this bewitching maths and beyond most 5-year olds but seriously people how can this be ignored in a platform decision?

How to choose the Platform for your new application
IBM has an approach called Fit For Purpose that looks at ALL of the factors above plus a number of others and proposes a decision tree for platform selection.  However let me propose an alternative method.  You made a decision about how you got to work today, classic Jaguar (hope that makes up for my earlier error), helicopter (Larry Ellison again), hovercraft, train, car or push bike or some combination.  Your decision probably included variables such as length of journey, time available, weather, cost etc.  Apply this same principle when you next choose an underlying platform for your application and remember that other platforms apart from x86 are available...