Go take a look at what Bill McColl has written. He's got a bee in his bonnet about real-time HPC. Trutthfully, while I think what he calls real-time HPC isn't there yet. I think we need another generation or three developmental cycles before its good and ready. ALL the HPC architectures that I have been briefed on have nontrivial issues with data movement to and from the machine itself. If you want real-time, you need low latency, uber double plus fat pipes to EACH node. If you say the words cluster and beowulf, I am going to have had my day made. I need a very good laugh.
Frex, the climate guys gave us a presentation, a little dated now but not a bad data point, where they wanted a 500 Petaflop (sustained with their code) machine for their ideal. We're working on single petaflop machines with the DARPA HPCS program. To do real-time, multipermutation climate sims seems out of reach for quite a while.
Plus, many of you will take issue with the 'generations' that Mr McColl has for the Internet. I am a relative youngin compared to some of you online and even I predate the 'internet generation 1'.
At any rate, what do my readers think?
Frex, the climate guys gave us a presentation, a little dated now but not a bad data point, where they wanted a 500 Petaflop (sustained with their code) machine for their ideal. We're working on single petaflop machines with the DARPA HPCS program. To do real-time, multipermutation climate sims seems out of reach for quite a while.
Plus, many of you will take issue with the 'generations' that Mr McColl has for the Internet. I am a relative youngin compared to some of you online and even I predate the 'internet generation 1'.
At any rate, what do my readers think?
YEP-- his 'generations' are somewhat accurate if we're discussing the web... but CRAP if we're actually discussing the internet itself... which predates the world wide web and the first browsers by at least a decade or more....
ReplyDeleteFrom the mid 80's to the early 90's I mostly remember the intrnet in the abstract.... reading ads for compuserve and aol in computer magazines and occasionally getting a vicarious taste of things through a friend or relative who was starting to make a habit of dialing into bbs bulleting board and simple text messaging systems.
In 92' I entered college and got my first taste of e-mail-- and while it seemed 'cool' it was severely limited-- you had to log directly into the school machines (and really know exactly what you were doing to access tcp/ip through your own machine) and could only send or receieve from others within the same intranet (no sending .edu messages to aol accounts or vice vers). Beyond email things again 'looked cool' but were actually fairly scarce.... gopher, muds, and mosaic-- (which, while having search engines like yahoo and being gui still lacked speed or any kind of multimedia (jpeg/mp3/avi) apps.
ok-- now back to McColl's hierarchy
---generation 1/2-- this distinction is sheer crap because the two essentially happened at the same time-- at least i know i was introduced to search engines at the same time I was first shown mosaic and that had to have been back in 92/spring of 93 at the latest.....
better distinction would have been gen 1-- text
gen 2... multimedia (around 95/96)
gen 3. -- higher speed dial up and the first signs of commercialy viable/commonplace broadband--- 97/98
*McColl's gen 3/4-- while these are 2 separate activities their arrival upon the net and especially the common mediascape beyond then net essentially happened at the same time (while things that could for all intents and puproses be called blogs started around 2000-- the explosion of social networking sites making blogs seem ubiquitous occurred around 03/04 just abut the same time podcasting and mashups were first being toyed with..
what he calls GEN 5 is basially just a dumbed down explanation of WEB 2.0 technologies/interfaces...
anyway, that's my rambling nostalgia/criticism of his '5 stages of the internet'