Date: Tue, 26 Sep 2006 06:57:34 -0400
Reply-To: ben.powell@CLA.CO.UK
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: ben.powell@CLA.CO.UK
Subject: Re: PC SAS vs. Mainframe SAS
Not exactly...
There is an order of magnitude difference in the number of servers required
(and to be administered) in a 2p single core infrastructure compared to 4p
multicore. Is 1/8th of a grid still a grid - 1/8th of a cluster still a
cluster? Bear in mind in 2008 we should have oct-core x86 in easy 32p
configs: that's the equivalent of a 128 blade cluster in one or two boxes.
(Where does that leave the interconnect business?)
As such the number of apps expected to run on single or a couple of boxes
increases massively. If a large chunk of your enterprise runs on a single
box what are going to want: linux or IBM?
I wouldn't rule out MF just yet, although the competition as ever is strong!
Rgds.
On Mon, 25 Sep 2006 12:25:16 -0600, Alan Churchill <SASL001@SAVIAN.NET> wrote:
>You're making my point Ben...
>
>This is what I think will be needed for not only today's large computational
>problems but will also be important as we move more toward SaaS models.
>
>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
>ben.powell@CLA.CO.UK
>Sent: Monday, September 25, 2006 1:25 AM
>To: SAS-L@LISTSERV.UGA.EDU
>Subject: Re: PC SAS vs. Mainframe SAS
>
>Bear in mind also that clusters have historically been comprised of 2p
>servers. Migration to 4p quadcore systems reduces a 8 server cluster to one
>machine. So we can add widespread adoption of 4p multicore servers as an
>additional threat to MF, and one that is far simpler to administer than a
>cluster, and cheaper to licence software for, or hardware glue, than grid,
>
>Rgds.
>
>http://www.ixbt.com/editorial/amd-guiseppe-amato-conf/105-enl.png
>
>On Mon, 18 Sep 2006 17:09:21 +0100, Ben Powell <Ben.powell@CLA.CO.UK> wrote:
>
>>"the death of the MF may be here"
>>
>>They've been saying that for 20 years :-)
>>
>>It may be true - it would be useful to see uptodate ROI. How many SAS
>installations are running on clusters and/or grid? Surely the admin for
>dozens of interconnected little machines is more of a headache than for one?
>Above 16 or 32 x86 cores you generally need non-standard interconnects
>and/or software glue to make the thing work. Whereas mf is otb. Competition
>then comes from SGI, Cray, etc, which is not PC,
>>
>>Rgds.
>>
>>
>>-----Original Message-----
>>From: Alan Churchill [mailto:SASL001@savian.net]
>>Sent: 18 September 2006 17:01
>>To: Ben Powell; SAS-L@LISTSERV.UGA.EDU
>>Subject: RE: PC SAS vs. Mainframe SAS
>>
>>Ben,
>>
>>The utilization issue is the whole point of grid computing. This is
>>already
>seen at the places I mentioned where CPU utilizations are certainly not at
>1%. Many of the benefits of mainframes can be found in clusters. Also,
>clusters can be upgraded much easier and are probably less prone to outages.
>>I doubt that the MF will hold out any advantages over a cluster
>>approach
>and hence why I think the death of the MF may be here.
>>
>>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
>ben.powell@CLA.CO.UK
>>Sent: Monday, September 18, 2006 9:44 AM
>>To: SAS-L@LISTSERV.UGA.EDU
>>Subject: Re: PC SAS vs. Mainframe SAS
>>
>>Wasn't it the arival of the internet that foretold the death of the
>mainframe? This was apparently disproven when IBMs mf sales jumped in 2003
>in part as a reversion to quality in the wake of the tech implosion. There
>were several articles on this at the time on the topic of "death of mf: much
>exagerated" except I can't seem to dig them up and all their roi arguments,
>comparing mf, pc and unix.
>>
>>I think its a slightly misconstrued argument to compare a single user
>experience in pc and mf. Of course the pc is going to be more satisfying in
>the majority of cases. But a mf purchase is not based on the needs of a
>single user but on 50-5000+ users, and not one app but all their apps. The
>minimum guaranteed service level of mf is also far higher than for pc - to
>spec pc support to similar levels requires not 1 or 2 highly trained support
>engineers but a whole army of it support.
>>
>>Couple that with security weakness and high rates of component /
>>software
>failure in pc, with little or no redundancy unless still more expensive
>solutions are built in and you can begin to see where the guaranteed uptime
>and centralised cost/resource model of mf begins to make headway, even
>today.
>>
>>Having said all that, I don't know since I can dig up the original
>>debate I
>am referring to, if this is all still the case, or if the balance has indeed
>swung massively in favour of pc. How are mf sales bearing up?
>>
>>Bear in mind powerpc processors from ibm already run quad core, which
>>would
>suggest some mf solutions should already be processing power competitive.
>>
>>Last but by no means least there is also the utilisation arguement. The
>>mf
>is generally configured to be running at close to ideal levels of
>utilisation so that little resource is wasted. This is rarely the case for
>pcs where 99% of the time the processor is virtually idle.
>>
>>I believe the article I'm referring to perhaps from mainframe month is
>"classic" and would be grateful if someone could post a link.
>>
>>As for whether google could run on mf rather than cluster, they must
>>have
>been approached by big blue, but I doubt they'd make the cba public,
>>
>>Rgds.
>>
>>
>>
>>On Mon, 18 Sep 2006 06:13:56 -0600, Alan Churchill <SASL001@SAVIAN.NET>
>>wrote:
>>
>>>Sander,
>>>
>>>
>>>I disagree. There is no way that a mainframe is stronger than a grid.
>>>A mainframe can be part of a grid and hence is a subset. As an
>>>example, a mainframe cannot do search queries faster than Google.
>>>
>>>The old comparison of a PC and a mainframe cost-wise is out of date.
>>>Today's PCs can be extremely safe and maintainable, they are fast in
>>>every measure, and they provide a much better user experience. Wall
>>>clock time, my guess is that the jobs will also run faster. With
>>>quad-core PCs available by this Christmas running virtualization, you
>>>could have an 8 way on the desk, running RAID 5 with 500GB of storage,
>>>and
>>4 GB of RAM for less than $5000.
>>>That is a big enough machine to handle the vast majority of SAS
>processing.
>>>Plus you can play Quake at ~200 frames per sec as a bonus ;-]
>>>
>>>
>>>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
>>>Sander Burggraaff
>>>Sent: Monday, September 18, 2006 4:07 AM
>>>To: SAS-L@LISTSERV.UGA.EDU
>>>Subject: Re: PC SAS vs. Mainframe SAS
>>>
>>>In all the answers people have given there is still missing one
>>>question: What are your programs for and what are the possibilities
>>>for running them Eric? Are they just typical batch programs or do they
>>>have to be run on demand?
>>>
>>>Mainframes are still stronger machines than any pc or grid. There's no
>>>doubt about that. Mainframes get updated and upgraded just as much as
>>>any other hardware. The question is how many people have to share that
>>mainframe?
>>>
>>>>From a developer's point of view a windows version is preferable
>>>although the ISPF editor isn't bad when you set it to 32 lines,
>>>program highlighting, parenthesis highlighting (a great option I have
>>>yet to find in other environments) and have something like TSO Plus
>>>which lets enable you to switch between screens. I 'grew up' on a
>>>mainframe so I feel quite at home. I've had colleagues complaining
>>>that they couldn't follow what I was doing because I was pushing
>>>buttons faster than the screens were showing ;-) But developing on a
>>>windows machine or another GUI environment is easier because you can
>>>run little pieces of code, change macro variables on the fly, etc., etc.
>>>
>>>One big advantage of the mainframe is it standard back up capabilities.
>>>The companies I have worked for so far always have had back ups made
>>>every night. That means if I screw up I am able to retrieve data or
>>>code from the day before usually within seconds. On pc or a
>>>non-mainframe network it usually means that you have lost the data or
>>>code or that someone has to insert a tape somewhere so that you have
>>>to wait days before you get your restore.
>>>
>>>Contrary to what many people believe there isn't a real price difference.
>>>The cost for mainframes seems high but many people forget that a pc
>>>probably costs just as much per person if you want to have all the
>>>safety and maintainability features a mainframe offers.
>>>
>>>So how big and important are your jobs Eric? If it were up to me I
>>>would choose to develop on pc and then port the programs to mainframe
>>>where they could be run in a batch. But that's probably out of the
>>>question ;-)
>>>
>>>With regards,
>>>
>>>
>>>Sander
>>
>>
>>***********************************************************************
>>***** The contents of this email and any attachments are confidential
>>to the
>intended recipient.
>>They may not be disclosed to, used by or copied in any way by anyone
>>other
>than the
>>intended recipient.
>>
>>Whilst any information and/or any opinion given is believed to be
>>correct,
>it is not intended
>>to constitute legal advice; you should seek specific legal advice as
>appropriate.
>>
>>Please note that CLA does not accept any responsibility for viruses and
>>it
>is your
>>responsibility to scan or otherwise check this email and any attachments.
>>***********************************************************************
>>*****
| 
