Mainframes and nuclear submarines
Nuclear submarines and mainframes have something in common. The customer base is fixed, your customers base has more than enough power at the moment, and you have difficulties to aquire new customers.
The comparision with submarines isn´t that far fetched than it sounds. You need very special know how for welding the hull. But, how do you keep this knowhow when your only customer has more than enough of your ships. Most of the knowledge isn´t really usable in non-military products. But the navy have a problem: You need new ones in 20-30 years. With mainframes is the same: A old-technology based cash-cow with a good customer base with high exit costs (as a customer told me: “I know it´s expensive, i know you systems are faster. But you can´t pay me enough to take the risk of migrating this essentially 30 years old code.” Honestly i wouldn´t do it either: The programmers of Version 1 to 2 already dead, the programmers of version 3-4 short before retirement) As there isn´t really a market for new systems, you can´t really fund the development of a next generation. Linux on Mainframe not really a solution. Okay, it´s a solution, but not for the problem, it´s touts to solve. It´s a effort to create demand for mainframe-MIPS thus increasing license revenue thus pulling new money into the concept of mainframe. As mainframes are only in usage in commercial enterprise systems the suffer more than other plattforms from the non-elastic demand of computing power in enterprise computing. Okay, the navy solve this problem by giving subsidies to those specialized ship builders, giving them “research assignments” or stretch updates over years that would be possible for less money in shorter time when done at once. With mainframes it´s somewhat different. Besides of their incredible expensive maintainance you can´t tell an enterprise customer to subsidise your business. For this reason many customers stay away from Mainframes as far as possible. No one would use a technology for new deployments. For example: Do you ever heard of a company using a mainframe for their customer internet portal? Besides of the ones with a press release? Would you really use a technology with the possibility of shortages in know how and personal in 5 to 10 years. For this very reason they try to sell their systems in the far-fetchest regions of the far-fetchest reasons. They need new blood in their customer base to keep their highly profitable customers in Europe and US. In a way those customer would be the life support for the systems running the skeletons in the migration closet of a number of CIOs. But the nice thing: IBMs attempt to sell their mainframes to China or Russia proofs, that this systems can´t be that powerful. Or do you really believe, that export regulations would allow IBM to export a system powerful enough to really consolidate hundreds of fully loaded linux machines to Russia or China ? ;)