Mainframe maker IBM and its few remaining peers that sell their own mainframes are probably cheered by two recent surveys of mainframe shops, one that came out of market researcher IDC and the other out of mainframe software maker CA. But maybe not. IDC polled some 300 IBM mainframe shops and kicked out a study with the …
A very nervous group, there
I've been telling people for over 15 years: the primary focus of mainframe people is finding ways to keep mainframe people employed. Given a task, they'll _never_ find a non-mainframe solution. I saw this at my previous employer, where the company president told them to implement company-wide email (circa 1992). They spent $85K on a package and another $60K/year for a disaster called EMCC/TAO which was slow, lost mail, crashed the clients constantly, and had to be restarted on the mainframe daily to keep from crashing it (memory leak). This was after I recommended a solution for a couple of thousand $. EMCC/TAO was so bad that our managers and engineers used modems and their own ISPs with Hotmail, Yahoo, or AOL accounts so they could communicate reliably with customers. They seemed to go out of their way to sabotage any non-IBM networks, like the NetWare servers I supported.
In the words of the Dilbert cartoon,
Here's a nickel kid, buy yourself a real computer**
**corrollary: licences sold separately.
Main frame creates lockin
Sure there is an argument to say it is cheaper to maintain as it is all one, but it doesn't really fly in the real world. It is better to build a varied interoperable network of computers that you can turn into a huge cluster if needed, that's a far better solution.
It is all going to smaller and many, not larger and one, it is more cost effective for more scenarios.
California may not have moved off mainframes, but that is not a testament to mainframes that's a lock in, and it will cost them more in the long run. Though the longer they can hold off, the cheaper it will be to replace, so at some time the sweet point will be found and off they will go to micro and distributed world, joining the rest of the human race.
Legacy software is key
To the survival of the mainframe. The only reason those people keep mainframes running is that they have 30 year old applications that no one dares to rewrite in a cheaper platform. Period. All those billions of clunky COBOL code, sometimes undocumented, and sometimes even without source code, require an up front investment in replacing that no one is willing to pay. Specially in the current climate that focus only on next quarter profits and annual bonuses.
@AC Not much of an argument
Not much of a reasoned argument against the mainframe - the people who work on it try to solve every problem using it... Errr well I find Java ppl will try and solve every problem using Java, C++ ppl the same, the DBAs will want to do everything in PLSQL. etc etc. There is also an example of a badly written mainframe app - I suppose ppl with little experience of the IT world might be surprised to hear of an expensive app created by seemingly intelligent ppl that runs on millions of quids of hardware and still doesnt work, not me, its a common story for all platforms and technologies! In fact the ACs example seems surprisingly cheap for a mainframe app :-)
I'm no MF apologist and have posted anonymously because I'm gonna say I think anyone who runs Java on Z is an idiot. Unless you really need the extreme levels of security provided the cost of running Java, even with zAAPs, is so high cheapo blades or even Power are a better option. Z runs DB and high IO workloads with unbelievable reliability and speed, what other platforms would you expect to run to 100% utilised? With the right workloads nothing is better than Z, but unsurprisingly sales ppl and stupid customers collide and we get Z where other platforms would be better.
At our shop we have a mainframe, a couple of Unix servers, and loads of Windows servers. Every couple of weeks we have a Windows server go down for 'emergency maintenance'. We are having a bad year if it happens once on the mainframe. Any wonder then that we are not rushing to move off the mainframe?