Have you been 'framed?
Mike Vinten asks whether mainframes still have a place
Most mainframe computer architectures were created in the 1960s, and the technology was in great demand until the advent of low- and high-end generalised servers in the 1990s. As the new servers developed into capable and reliable machines, did they make mainframes redundant? Is such a dated, high-priced, inflexible technology needed at all?
In the mainframe market, IBM is the principal player, controlling around 90 per cent of the market. Unisys and Fujitsu, which still have some dated ICL sites, are trying hard to protect their revenue. Last-minute negotiations in an attempt to retain machines even for a short while – during a modernisation or migration – will cost you a fortune.
Distributors of servers and storage will see a huge new market in the large mainframe estates, mostly controlled directly by the manufacturers. Partner share is small or non-existent. For example, IBM passes on 30 per cent of the revenue while holding 70 per cent for itself.
IBM further restricts things, because it is not cost-effective for it to sell to low-end customers. This short-sightedness creates a market where there are more companies selling migrations than mainframes.
Market numbers speak for themselves. In the UK, few IBM mainframe sites now remain, with only around 140 sites left out of thousands. Fourth quarter revenues in 2009, I have read, were also down 27 per cent on the previous year. For Fujitsu mainframes, the situation is much worse. The dated VME platform is only kept afloat via lock-in contracts, many with the government.
The immense range of open, integrated and flexible technologies does not sit easily with mainframe. The architecture, superb in its time, does not interconnect readily. As a result, it is just not worthwhile for hardware and software developers to target the mainframe market.
The autonomous IBM mainframe software division, a cash cow for Big Blue, is reluctant to reduce its prices more often than not, fighting its own hardware division. There is openly competition with third-party software providers that have added fuel to the fire by draining revenue from the mainframe market.
Cheaper options like Linux or UNIX operating systems on mainframe have not been a success. Many Linux engines have been provided for free but are hardly used. Most have compatibility and back-level issues if migration from other UNIX systems is involved.
Customers locked in with large development teams developing bespoke applications have large hidden costs. Mainframe statistics often show that very high-end price performance is excellent, but they miss out on the more complex and less automatic processes that wrap around the system costs.
Mike Vinten is a self-employed consultant at Discover1