Date: Mon, 30 Jul 2007 16:48:35 -0600
Reply-To: Alan Churchill <savian001@GMAIL.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Alan Churchill <savian001@GMAIL.COM>
Subject: Re: How much? [SAS-L]
In-Reply-To: <00cf01c7d2f8$7ebb8790$2039688e@MSG.COUS.UVic.CA>
Content-Type: text/plain; charset="iso-8859-1"
Cycling Windows servers down is definitely an issue. A lot of progress has
been made by Microsoft to reduce reboots over time but they still occur as
most of us are aware. Best to do it in cycles of machines. Expect further
progress in this area with Windows Server 2008 (Feb release).
The comparison is really not fair with 1 PC to 1 mainframe. The real
comparison is 1 datacenter running massively parallel PC clusters vs a
datacenter running a mainframe. Google or Microsoft datacenters vs
mainframes is more apt. How many transactions does Google serve up a day and
how fast can they serve them up...
Of course, my Vista machine that I purchase around Christmas will be 64 bit
and will have loads of RAM.
Finally, the RAS argument I think is a bit more difficult with the massive
architectures in place from Microsoft and Google. They are highly, highly
redundant. With Amazon, the 3 are literally running over 1 million servers
in their datacenters. That is a level of hardware redundancy that, I
believe, is not replicated anywhere.
Alan
Alan Churchill
Savian "Bridging SAS and Microsoft Technologies"
www.savian.net
-----Original Message-----
From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of Melvin
Klassen
Sent: Monday, July 30, 2007 4:25 PM
To: SAS-L@LISTSERV.UGA.EDU
Subject: Re: How much? [SAS-L]
> I'm curious - why do you *want* to price a mainframe, when
workstations and servers are less expensive ...
Be sure to measure the TCO (Total Cost of Ownership) correctly,
i.e., how many Microsoft Windows servers do you have to take offline and
security-patch
on the second Tuesday of every month? How many programmer-hours of lost
productivity are there, per month, in doing the patching?
Don't compare apples to oranges (or off-the-shelf workstation-quality PC
hardware
to highly-reliable mainframe), before you say "B is cheaper than C".
> and (in some but not all ways) more capable?
> Are you doing something that requires huge numbers of
> simultaneous users, or requires software not available elsewhere?
>
> I don't know how much faster (or slower) mainframe disk storage is
than
> workstation storage, but it's certainly more expensive.
The local mainframe uses "striped I/O" (each byte fanned-out over
multiple hot-swappable hard-drives),
and dedicated controllers, with large I/O buffers, to make I/O much
faster on the mainframe,
when compared to a single PC-class hard-drive.
As in drag-racing, speed costs money -- how fast do you want to go?
> Ditto for memory, except that you also can't use as much (for SAS,
anyway) on a mainframe.
Nonsense. The nearby mainframe uses 31-bit addressing (2GB of
virtual-RAM per user).
Microsoft Vista seems to like 1GB per PC. Yes, it's an "old" mainframe;
any replacement
would use 64-bit addressing, while continuing to run "legacy" 24-bit and
31-bit applications,
unlike SAS 9.1 for Windows 2000/XP, which stumbles when confronted with
Windows Vista.
If you like RAS (Reliability, Availability, Serviceability), you'll like
the mainframe.
