The rest of the discussion (here and over on the mainframe blog) is a fairly childish “tis so, no tisn’t” argument between Tim and Charles about costs. Both sides are somewhat ridiculous because they’re comparing pieces of string. Mainframe, Windows, Linux and UNIX each have valid and viable roles in practically every significant enterprise on the planet today. You’re not going to replace my PC with a mainframe or even a mainframe terminal. It’s fantasy to think I’d even want that. On the other hand, I really don’t want my bank account details stored on a commodity SATA drive connected to a commodity Windows PC in my local bank branch.

IT stuff is expensive no matter who you buy it from. Could it be more efficient? Well sure. But a little reality in the discussions would be more productive than strawman arguments that prove nothing - especially when both sides are so coy about what things really cost.

One fundamental reason for HIS is because it is always preferable to run an instruction off the mainframe as opposed to on it, given a choice. The reality is the mainframe is the most expensive place in the world to run code and getting more expensive in relative terms. Welcome to the land that Moore’s Law forgot. Last time I checked (and I haven’t been paying attention for a while), the mainframe was three orders of magnitude more expensive per MIP than x86 and was falling further and further behind. You’re crazy to put any new workload on the mainframe. This is why IBM is always peddling the fad of the moment on the mainframe to see if they can hoodwink people to maintaining or even increasing the workloads on their mainframe. Run Linux and Java on the mainframe, they say, never mind the fact that these cross platform approaches should drive people to the lowest cost hardware, not the highest cost. This week they’ve probably got a sales pitch for how you can Second Life on the mainframe. Anything to keep the mainframe annuity flowing in.

IBM is welcome to disprove this by actually publishing industry standard benchmarks for the mainframe, but for some reason they have not been willing to permit any apples to apples comparisons for years.

Whatever their rationale, customers seem to prefer HIS. The last time I checked about 60% of mainframe installations have an instance of HIS or precursor running.

For all the ways IBM has pitched the “mainframe renaissance” in recent years, the reality is the mainframe base continues to shrink and IBM continues to jack prices for the customers who remain. Perhaps a mainframe blog is just what it will take to turn this around in the face of terrible economics, limited investment, a dying ecosystem and a graying talent pool that isn’t being replenished.

To put this in perspective, we’ve just been through the biggest computing buildout ever in the last decade with the Web, and the mainframe is nowhere to be seen. Even when IBM has tried to pay customers to run portions of their web sites on a mainframe, they have failed. There is a lot of inertia to the installed base, and IBM has been very active in prolonging it as much as possible, but it is a dead end.