Notes on two presentations by Gordon Bell ca. 1998
|
From: Rick Jones on 10 Feb 2010 17:37 Robert Myers <rbmyersusa(a)gmail.com> wrote: > Andy Glew has said that a supercomputer is nothing but a big, multi- > tiered switch. That is certainly what supercomputers have come to > be, and I don't want to give the impression that I think even the > lame interconnects we get are necessarily trivial. Does it even have to be a "supercomputer" to look like a big multi-tiered switch? So, a system near to my paycheck - HP Superdome - a "cell based" (no, not *that* cell thank you very much :) system - two processors on an FSB with an agent chip that speaks to an interconnect fabric. I suspect it would be considered a "real" computer produced by "real" architects and developers. (those feeling snarky are encouraged to send your snarks in direct email) However, if I look at the block diagrams, and squint only a little, it doesn't really look that different from those two processors being small systems connected to a switch called "agent" which itself is connected to other "switches" with other names that then lead to other small systems. Now, this "switched network" isn't passing Ethernet frames, it is passing coherence protocol, but it certainly does seem to look rather the same, with different constants. rick jones -- oxymoron n, commuter in a gas-guzzling luxury SUV with an American flag these opinions are mine, all mine; HP might not want them anyway... :) feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...
From: Eugene Miya on 10 Feb 2010 17:43 In article <hksvhp$ru5$1(a)usenet01.boi.hp.com>, Rick Jones <rick.jones2(a)hp.com> wrote: >I may share what I perceive to be your disdain for calling a bunch of >systems connected via a network a "supercomputer" but in the case of >Avatar, if what I've heard is correct, it was rendered on hardware >enough of which was "general purpose" that if the folks who bought it >wanted to, could break it up and sell the blades/chassis to people who >wanted to do general purpose stuff. So I'm not sure how much of what >was used to render Avatar would be "one-off." The problem for your (say HP's side) is that while the hardware was state of the art when say you sold it to a firm: a house, a producer, when the film is done, you will be ready to sell the next upgrade. This was a problem which killed Osborne as a firm. At the higher level that's fine for 1 producer/director. Studios coordinate multiple films or TV series, attempting to keep all secret. You have to have pipelines of multiple graphics units (groups) juggling in the air. I was amused by a screening room I was invited into in Redwood City (they have 2). They wanted military/govt grade crypto between there and their contacts in Los Angeles suburbs. These are not investments Studios until recently have been willing to make (I showed a brief bit of footage which was done for Alien which also appeared in Bladerunner, the fan, who had dinner with Scott recently was surprised). It's called show business for a reason. The problem, like publishers, is that they (the MPAA) don't the pace of Moore's Law. -- Looking for an H-912 (container).
From: Rick Jones on 10 Feb 2010 17:53 Eugene Miya <eugene(a)cse.ucsc.edu> wrote: > The problem for your (say HP's side) is that while the hardware was > state of the art when say you sold it to a firm: a house, a > producer, when the film is done, you will be ready to sell the next > upgrade. This was a problem which killed Osborne as a firm. > At the higher level that's fine for 1 producer/director. Studios > coordinate multiple films or TV series, attempting to keep all > secret. You have to have pipelines of multiple graphics units > (groups) juggling in the air. I was amused by a screening room I was > invited into in Redwood City (they have 2). They wanted > military/govt grade crypto between there and their contacts in Los > Angeles suburbs. These are not investments Studios until recently > have been willing to make (I showed a brief bit of footage which was > done for Alien which also appeared in Bladerunner, the fan, who had > dinner with Scott recently was surprised). It's called show > business for a reason. The problem, like publishers, is that they > (the MPAA) don't the pace of Moore's Law. The don't what the pace of Moore's law - like, keep-up with? Anyhow, since you mentioned TV series and films - and since it is a "business" - will the Penguins of Madagascar animated series actually be using "state of the art" pipelines or will they have gotten "hand-me-downs" from say the production of Madagascar: Escape to Africa? I'm sure a systems vendor would *love* to sell them the latest and greatest, but ginve that is is called show business are they really going to be buying brand spanking new equipment for every project? rick jones -- a wide gulf separates "what if" from "if only" these opinions are mine, all mine; HP might not want them anyway... :) feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...
From: Robert Myers on 10 Feb 2010 18:25 On Feb 10, 5:37 pm, Rick Jones <rick.jon...(a)hp.com> wrote: > Robert Myers <rbmyers...(a)gmail.com> wrote: > > Andy Glew has said that a supercomputer is nothing but a big, multi- > > tiered switch. That is certainly what supercomputers have come to > > be, and I don't want to give the impression that I think even the > > lame interconnects we get are necessarily trivial. > > Does it even have to be a "supercomputer" to look like a big > multi-tiered switch? > The term supercomputer is now nearly meaningless, so there's no point in haggling over it. ;-) > So, a system near to my paycheck - HP Superdome - a "cell based" (no, > not *that* cell thank you very much :) system - two processors on an > FSB with an agent chip that speaks to an interconnect fabric. I > suspect it would be considered a "real" computer produced by "real" > architects and developers. (those feeling snarky are encouraged to > send your snarks in direct email) > I'll use your mention of Superdome to return to a subject presumed dead, which is Itanium: http://www.tgdaily.com/networking-brief/48339-intel-itanium-outsells-amd-opteron "Intel kicked off its Itanium presentation today by saying the Itanium's system revenue since the introduction of 2001 has crossed the $5 billion mark. That outsells total sales of AMD's Opterons." "And, according to IDC Itanium sales surpassed all of SPARC sales for the first time since the introduction of the Itanium." While some will undoubtedly think my introduction of Itanium into the discussion gratuitous, those people probably never have been much interested or understood what I have to say on the subject; viz, When it comes to moving things around at any level, learning from the past behavior of given software is at least as promising an avenue for development as hoping for miraculously better hardware. If you know well ahead of time where things need to be, you can be a lot smarter about time and energy costs of data movement. I know that's not a particularly original insight, as it has been understood and pursued with varying degrees of focus since the earliest days of computing. So, in direct response to your comment, maybe "big multi-tiered switch" is the paradigm for everything but personal computers. Until anyone figures out another way, maybe it would be better to spend all that money trying to make software less clueless. Robert.
From: Robert Myers on 10 Feb 2010 18:30
On Feb 10, 5:34 pm, eug...(a)cse.ucsc.edu (Eugene Miya) wrote: > > My cell phone's software still has problems, I have to get around to > getting a new laptop, but I don't see a whole lot of problems with that. > It's a wide field. > I owe my awareness of comp.risks to you. I just had to edit the registry by hand just to "renew" the expired license of my security software. I had to remove dead registry keys that made the already-installed security vendor's software think that a previous security vendor's software was still there. I guess you can call that working, if you want. I can just imagine advising a puzzled friend to invoke regedit. Robert. |