John Campbell wrote:
[GRAIN TYPE="SALT" MODE="Stand Up Philosopher"]
A long time ago in a website far, far away, I once sent a note to the
fellow talking about mainframes.
In fact, some years after I wrote the original author a note and then
forgot about it, I discovered that my reply had somehow become holy writ in
the definition of "what is a mainframe"... all this with minimal
experience on any IBM mainframes (most of my mainframe time was on Xerox
Sigma-9s, UNIVAC-1100s and the like) but, given a fair quantity of eclectic
experiences, I pointed out what *I* saw as the key differentiators.
Me, a pundit? Surely an accident!
Now I have to qualify what I say here because of my e-mail address. Sadly
(for me, at least) I have *minimal* exposure to mainframes and even p5
systems aren't common for me, despite being an AIX/Linux/Network admin.
The point isn't that a mainframe CPU isn't much faster than an Intel CPU
(or Power5) but that it is *reliable*. Each CP resides in a system
designed for non-stop operation, too, which can be serviced *WHILST
RUNNING*.
The term "mainframe" has been around for some years.
Remember "Tandem Non Stop?" Tandem's big selling-point was reliability
(fault tolerance) through redundancy. As best I can recall I went to a
presentation in the early 80s, when we had Amdahl 43xxs and IBM's
mainframes were 3080 series. I think we had some dodgy disk drives, but
otherwise reliability was fine.
My best recollection of mainframe reliability goes to IBM's early
S/370s. We had /135s in Brisbane, Adelaide and Perth, and /145s in
Sydney and Melbourne.
I was based in Canberra, and had occasion to visit Melbourne several
times and Brisbane once.
I was present when the Melbourne 145 was down (hardware) and required
IBM's urgent attention once, maybe twice.
I was present when Brisbane's 135's printer-keyboard was down making the
system essentially unusable, also requiring IBM's urgent attention.
In Sydney, we had regular "software" errors resulting on 106-F abends in
various tasks including initiators and writers, and system wait states
(code 028 comes to mind) and other errors. All arose from read errors
(NRF) from IBM disks on IBM channels. IBM software said it was hardware,
IBM hardware said it was software. The problem didn't go away until we
upgraded the DASD with Memorex drives.
_I_ don't think "mainframe" means "reliability," or that it ever did.
I don't dispute that current IBM mainframes are any less reliable than
everyone says they are, but to my mind "mainframe" goes to the
architecture of the box and (maybe) the software found running there.
Yes, a completely different animal-- an Intel CPU is *not* easily
equivalent to a CP for a "z", but, then, I don't see people wanting to run
[EMAIL PROTECTED] on a z9 either unless they want to exercise the floating point
hardware (mind you I'd find watching this to be entertaining, to say the
least).
I have not heard of any failed Intel or AMD CPUs in a very long time.
Accompanying system components such as RAM, disks, NICs, yes, but not
the CPU itself. The systems I use are built to be cheap; one can have
greater reliability for a greater price. I imagine that for the price of
a mainframe one could also have a fairly reliable IA32/AMD-64 system.
So, yes, the author of the blog on ZDnet was, in my opinion at least,
comparing apples and oranges... simply because the hardware platforms are
optimized for two different operating regimes... and they're not easily
compared.
I have *no* current experience to base my prejudices on, of course...
Just remember:
#1: A mainframe is designed to provide maximum *reliable* single-thread
performance (per CP) simply because a lot of the business workload (merge
phase of sortation, for instance, or balancing a b-tree after an insert)
needs it and errors can't be tolerated.
Seriously, I don't believe that a typical IA32/AMD-64 system would do
this less reliably. A current zSeries may well do it faster though.
#2: A mainframe provides maximum I/O connectivity. Granted, as we move
towards fibre channels we're getting away from this differentiator. I wish
I had a Shark ESS to play with...
#3: A mainframe excels are maximizing I/O bandwidth across *all* devices
in order to minimize choke-points, especially when dealing with databases
(and their key structures).
How do current big Suns, IBM pSeries & iSeries (does HP still have
something in this market) compare here?
How about Univac's mainframes?
--
Cheers
John
-- spambait
[EMAIL PROTECTED] [EMAIL PROTECTED]
Tourist pics http://portgeographe.environmentaldisasters.cds.merseine.nu/
do not reply off-list
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
