On Apr 7, 2013, at 8:49 AM, Guy Sotomayor <ggs at shiresoft.com> wrote:
On Apr 7, 2013, at 3:16 AM, MG <marcogb at xs4all.nl> wrote:
On 7-apr-2013 7:55, Dave McGuire wrote:
*laugh*
I'll tell IBM to "get right on that". I'm sure they'll value your
sage
business advice.
By the way, you should ditch the Cold War mentality at least. Not
everybody who doesn't blindly kiss the /behind/ of IBM (like you)
is a 'traitor'.
*sigh*
I wasn't going to jump into this but...
For those that say the mainframe (specifically IBM 360/370/390/z-series architecture)
is dead, I would say not. Not only are they suited to do traditional
"mainframe" tasks
but they're *really* good at virtual hosting. You know the virtualization stuff
that's been
touted by the likes of VMWare, Microsoft, etc on Intel platforms for the last few years.
Well, IBM's been doing that since 370 days (early 1970's). They do
virtualization better
than anyone...almost no IBM shops run the OS on the "metal". They're run
inside a VM.
It's just that good.
Linux has been ported to the z-series. As an experiment, IBM wanted to see how many
Linux VM clients could be run (basically to see what it would be like for a hosting
company).
They stopped at 42,000 Linux clients running on a *single* z-series. They stopped not
because the performance was unacceptable or because they ran into any limits...they got
tired of adding VMs!
It's an architecture that will soon be 50 years old. The original 360 architecture
(which was
not only the ISA but also how I/O was interfaced) has been surprisingly resilient.
I cast suspicion on anyone who says something is "dead". Not only have I been
hearing
that the mainframe is dead since the late 70's, I've also been hearing that the
disk drive
is dead for about the same period of time. How's that going?
Oh, and here's some more on the zSeries (copied from wikipedia):
A direct comparison of System z servers with other computing platforms is difficult. For
example,
System z servers offload such functions as I/O processing, cryptography, memory control,
and
various service functions (such as hardware configuration management and error logging) to
dedicated
processors. These "extra" processors are in addition to the (up to) 101 CPs per
frame. System z cores
include extensive self checking of results, and if an error is detected the server retries
the instruction.
If the instruction still fails, the server shuts down the failing processor and shifts
workload, "in flight," to
a surviving spare processor. The IBM mainframe then "calls home" (automatically
places a service call
to IBM). An IBM service technician replaces the failed component with a replacement part
(possibly even
a new processor book, consisting of a group of processors). With System z9 servers, the
technician
installs the new book and removes the old one without interruption to running
applications. (Note that
mainframes have reported MTBF figures of 20-50 years). Similar design redundancies exist
in memory,
I/O, power, cooling, and other subsystems. All these features exist at the hardware and
microcode level,
without special application programming. The same concepts can extend to coupled frames
separated
by up to 100km in a Geographically Dispersed Parallel Sysplex when z/OS is used.
As has been said by others, mainframes are used by organizations that care about data
availability
and uptime. BTW zSeries, the z stands for "zero downtime".
TTFN - Guy