MF folks would love an unlimited budget. Just like
all of us would. So what? Silly.

The question is, did they satisfy the user needs
based on the available money?

No.

Are the user needs crucial to the running of
the business?

Maybe. Depends.

The bean counters wanted reports. The MF programmers
might take a couple of days to weeks to produce the
report. The bean counter wanted a change. The cycle
would repeat.

A single bean counter can think up dozens of reports
every month. But the programmer can only produce 2 or 3.

So there became a huge backlog of IT projects to satisfy
the reporting requirements, with no end in sight. There
were never enough programmers or CPU cycles to come close
to denting it. It was the programmer manager's job to
say NO. "You must fund me more programmers and mainframes
in order to create your reports." Which of course rarely
happened unless the business was incredibly profitable.

PCs came in as Visicalc machines, not PCs. This
was on an Apple. Later they came in as Lotus 123 machines.
People did not think of them as computers, they were
task specific tools.

A bean counter with a PC could do his work, plus the work
of 5 programmers (based on previous productivity). So
of course nobody ever trusted the programmer manager
again.

Fast forward another 30 years. Wow. In that time there
have been tremendous leaps in the various non mainframe
compute ability. Unless the key issue is backwards
compatibility, there is NO reason to pay IBM 2MM for a
mainframe that a pair of high end Sun/HP boxes won't do
better, faster, as reliably, almost as securely.

I say as reliably because MF sysplex clusters
are no better than the high end Sun or HP clusters.

I say ALMOST as securely, because it seems that IBM really
does have some major hot stuff in the hardware encryption
security side, meaning even the memory if encrypted. But since
I have never experienced a company (not gov agency) that
cares at that level, the next level down with MAC (Mandatory
Access Controls) in the some the secure Unixii is "good enough".

As far as "user counts", though, the default MF model is
far better than the default Unix one. Unix people have to
work hard, in tandem with the application and database
designers when attempting to scale.

As far as high velocity batch processing, the mainframe SUCKS
unless you are willing to spend 10 times the amount compared
to a Unix system.

I've actively moving jobs off a NEW mainframe to Sun Solaris
and Intel Linux. Things that used to DAYS now take minutes.

So, what does a company do? "Train" their MF people to
become Unix people? I've got experience with this.

Most MF people I know have no concept of CPU time, disk
space, IO rate, etc. They know if they submit their
job some time later it might finish. That's IT!

They claim to understand CPU time, but then accept
wildly varying run times. They program in isolation,
but run shared, and rarely account for it.

They are so used to taking a long time, fighting for
resources, that they might let a job run for DAYS
that should go in minutes. So they expect the same
performance on the Unix box, not realizing that their
job aborted 2 days ago.

These people might make mediocre programmers anywhere,
but as MF programmers they were able to get away with
it. Put them on a Unix box and lose what little
productivity they already had.

The peope who do know the details are so immersed into
the box that they would rather kill before moving to
another environment. These are the hard core MF techs.
There is very little cross-over into other environments.
They spent years honing this knowledge. The last
time something showed any promise was OS/2, and look
what happened to that. They know PCs have horrible
stability issues (Windows, etc), and they think in
term of the MF vs everything else, which means that
the Unix systems are thought of as PCs as well.

The MF managers are used to being a major player. They
commanded budgets of 10s of millions of dollars JUST
for hardware. Plus all their systems programmers, operators,
tape monkeys, etc. And it has worked so far. Why would
they risk their career on any other box? So they fight it
tooth and nail.

Every dollar that does into a Unix project is 10 dollars
removed from the MF budget. Really. They do NOT like
this, especially when some young punk is pushing the Unix
side and looking good doing it.

Also, I've noticed a trend on MF people who pickup PC
skills end up using the PC just like any other PC user
would. They like it. It is SO much better than what
they are used on the mainframe. Using Excel, Access,
pretty report writers, etc. As opposed to Unix geek
who uses PCs. They are forced to, constantly coming
up with reasons to do things on the Unix/Linux box.

So there is yet another class war. The PC using MF
people just can't understand why they have to learn yet
another system, ie: Unix. Since the Unix system integrates
well (Samba, ODBC drivers, etc), it removes much of
the reason for a PC using MF person to then understand
Unix, which then means the really cool stuff like
piping never happens.

So they end up using the Unix box like anothe mainframe,
but are able to edit their programs via Samba shares.
It is LESS productive than the MF since they had a good
programming environment there, but they never bothered
to learn it under Unix.

I've been doing this 6 years in a MF/Unix shop. I have NEVER
seen a MF person move to Unix with any real success. The
really smart MF people ended up being managers of the Unix
programmers. They understand the power but not the
technology.