I've been mostly on the mainframe side of the shop 'forever' - just over
two decades. If I was a CIO/CTO, I'd use the mainframe for the big DBs,
and the big IO, which seems consistent with the thread. I don't think
I'd have my email server there, or any of the other ancillary functions
that seem so well suited to their own box. I wouldn't burden a big blue
machine with DNS. I would let RACF (excuse me - Security Server) in/on a
mainframe VM handle all the authentication for aps.
"System Programmer" has taken on a whole new meaning since my first
days. Now, if you know where the on/off switch is, you too can be a
sysprog. Some of the real MVS chefs out there have taken to Linux very
well, and the integration of additional applications goes as smoothly as
loading the latest 'PUT' tape from 'PID' (neither of which exist in
their original form anymore).
The big blockade is bucks. Mainframes have been around so long that the
cost of ownership issues are well understood. The smaller machines have
a much more diverse audience and expense base than can be easily
accounted for. I know this to be true as I do cost of ownership
determination and control consulting for a living. Pouring money into
small machines is relatively painless, like boiling a frog starting with
the frog in cold water. Expenses creep in gradually, and the frog
doesn't realize how much his entire system costs until he's cooked. Big
iron = big bucks from day one.
I wouldn't put additional Linux stuff on the mainframe just because I
could. Those MIPs/MSUs come at a high per unit price, and are better
spent where large amounts of data needs to be shared at high speed. The
little boxes are catching up, but why take a chance on affecting system
performance by slicing and dicing the machine into more and more pieces?
Even the most efficient hypervisor code takes a finite number of machine
cycles. It's a little more than just a task switch, dontcha know... Try
explaining to the president of the company/school/etc that your
mainframe needs to grow again - three years before you projected -
'cause you loaded it up extra VMs! If you have cycles to burn, have fun!
Just have a plan and budget ready to migrate applications that were put
on the mainframe 'cause you could'.
If I've misinterpretted what I've read, my apologies for the 'soap box'
treatment.
Mike Galgano
Mark C. Ballew wrote:
Here's the thing: We do have mainframe people in place, and it will
still be many years before they retire/get frozen for the year 2038.
So on the one side we have 390 admins, and on the other we have
people like me who do the Linux stuff.
This is a shop where we've always done things the mainframe way, and
when we can't use a mainframe we use AIX on the pSeries. So now that
Linux is the new kid on the block, it seems logical to put it on the
mainframe to these guys[0].
I think that Brian brought up a good point on processor scaling. We
aren't running Oracle, more like DB2, and my understanding is that
there are seasonal load spikes on the databases (student registration
and the like). A database sounds like it could handle the I/O, but at
some point we also generate reports[1], so would a 390 machine be
able to crunch the numbers on multiple CPUs compared to some Linux
servers? Admin work hour for work hour, which is the most efficient
assuming we have trained employees on both sides[2]?
Mark
[0]I didn't like Linux on the pSeries after doing extensive
investigation and IBM training
[1]Time to break out the Cobol
[2]Note that the UNIX and Linux people co-exist with the mainframe
people pretty well
On Aug 31, 2005, at 1:25 PM, James Washer wrote:
If you don't have a mainframe admin staff already in place, it's
hard to recommend Linux on 390, as the learning curve is steep.
If you do have such a staff, then linux on 390 is a beautiful thing.
Any Virtual Machine is great but 390 is the mother of all VMs.
If you don't need true mainframe horsepower (i/o bandwidth for
example) then you might like Vmware GSX or ESX on a big intel box..
- jim
On Wed, 31 Aug 2005 11:47:56 -0700
"Mark C. Ballew" <[EMAIL PROTECTED]> wrote:
This question might be targeted more toward James, but what do people
think about Linux on the Mainframe for deploying Linux? I'm looking
at using zSeries machines to run various databases under Linux, since
mainframes are often acclaimed for their i/o abilities.
Assuming money is no problem, what would be the reasoning behind
going for a normal Linux server setup with fail over compared to
using an underlying operating system such as zOS? What about a Linux
server set up running UML or Xen versus zOS?
Factors that I can think of off hand is that adding more Linux
servers is easier, and software compiled for i386/x86-64 would be far
more common, but for the software I plan to run, a zOS port would
certainly be available. Other things would be the physical labor
involved: that is something that will always be limited. There are
probably more people in the industry that know how to manage Linux
servers than Linux on zOS. zOS is also a "blackbox", so an admin
would have to rely on the vendor instead of training to fix a problem
in some cases.
Any one else have thoughts?
_______________________________________________
RLUG mailing list
[email protected]
http://lists.rlug.org/mailman/listinfo/rlug
_______________________________________________
RLUG mailing list
[email protected]
http://lists.rlug.org/mailman/listinfo/rlug
_______________________________________________
RLUG mailing list
[email protected]
http://lists.rlug.org/mailman/listinfo/rlug
--
"Whoever said the pen is mightier than the sword obviously never
encountered automatic weapons."
- General MacArthur
--
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.344 / Virus Database: 267.10.17/85 - Release Date: 8/30/05
_______________________________________________
RLUG mailing list
[email protected]
http://lists.rlug.org/mailman/listinfo/rlug
