And there will be both fixed costs and variable costs to any computing platform. Mainframes are generally characterized as having moderate fixed costs and low variable costs. Other systems often have lower fixed costs (individually -- collectively they can very easily be higher) and "unpredictable" variable costs.
This is such a big problem, this lack of understanding about fixed-plus-variable, so I thought I'd mention it again. This is very much like your electric bill, or your telephone bill -- especially with peak and off-peak pricing. With your electric bill, there's a base service charge, and then there's a variable charge (with a peak/off-peak adjustment). If you attempt to collapse everything into one cost number you can get into big trouble and end up with wildly perverse incentives. For example, let's suppose you double the number of CICS transactions (and associated batch) you are running today on your mainframe. Do your costs double? Well, your chargebacks would if you used a single number like "average cost." But no, your costs don't double -- nowhere close to that! For one thing, your operations headcount (salaries, benefits, office space, etc.) probably remains exactly the same. Your data center facilities charges (space, power, cooling, etc.) would also remain exactly the same (or near enough). Yes, your software and hardware charges would increase, but those won't double either -- still not even close. (Mainframe software prices, in particular, are strongly curved.) Here's another way to think of it: if fixed costs really were zero, why would anyone bother to consolidate IT when two companies merge? IT integration, sure, but consolidation? (Why close any data centers, for example?) Of course fixed costs matter. So why do so many people pretend they don't exist -- and adopt chargeback regimes which pretend they don't exist? As an aside, there are many people that say that mainframes are batch systems which provide free on-line service, or on-line systems which provide free batch processing. And what people mean is that the only factor that determines the total hardware and software price nowadays (in the era of WLC) is your monthly peak. (Hardware is largely part of the fixed price element, while software is a lot of variable and some fixed. But even the fixed components are determined according to something like your forecast annual peak.) If your peak is at stock market open, that stock market open behavior determines the hardware and software price. That's how it works in a shared-everything environment. With other types of IT infrastructure, prices likely increase when *any* peak at any time of day increases. For example, if the Internet application spikes to a new and higher peak at 6:00 p.m., then that increases your price because you have to increase computing resources allocated to that particular Internet application, even if your market trading application at market open is running on infrastructure that is largely idle at 6:00 p.m. You might have 1,000+ little mainframes rather than one, basically. OK, hopefully that wasn't confusing. :-) - - - - - Timothy Sipples IBM Consulting Enterprise Software Architect Based in Tokyo, Serving IBM Japan / Asia-Pacific E-Mail: [email protected] ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
