Comments on: Intel’s “Missing Middle” Revives “Attack of the Killer Micros” http://insidehpc.com/2010/09/15/intels-missing-middle-revives-attack-of-the-killer-micros/ HPC News Without the Noise for Supercomputing Professionals | insideHPC Sun, 19 May 2013 10:46:54 +0000 hourly 1 http://wordpress.org/?v=3.1.1 By: The missing middle, a marketing term, but a real problem « scalability.org http://insidehpc.com/2010/09/15/intels-missing-middle-revives-attack-of-the-killer-micros/#comment-255873 The missing middle, a marketing term, but a real problem « scalability.org Thu, 16 Sep 2010 17:05:26 +0000 http://insidehpc.com/?p=13551#comment-255873 [...] is being addressed by some (hint: the day job) in a variety of ways. An article at InsideHPC by Rich Brueckner gave it a good contextual background, in terms of historical trends [...] [...] is being addressed by some (hint: the day job) in a variety of ways. An article at InsideHPC by Rich Brueckner gave it a good contextual background, in terms of historical trends [...]

]]> By: Mark Hahn http://insidehpc.com/2010/09/15/intels-missing-middle-revives-attack-of-the-killer-micros/#comment-255737 Mark Hahn Wed, 15 Sep 2010 17:51:04 +0000 http://insidehpc.com/?p=13551#comment-255737 it's a little hard to see the point here. yes, you can do HPC on NOW, and you can "serverize" the NOW with IPMI and cages. so the take-home is mainly that cube people need displays, and it's cheap to attach a display to each node of the cluster? OK, sure - it's not that it doesn't make sense, but that it's not news. and it's slightly dubious in that it goes against thinclient/virtualization trends, and also assumes the workload is not network-sensitive. (I would claim that the latter is actually the _defining_ feature of HPC.) this is almost a spread-the-pain kind of thing: take servers out of datacenters and spread them around the human-occupied spaces. noise is mostly a design issue, and at least in climates where buildings have to be heated, you get some small green benefit. in other words, the main thing keeping servers in the machineroom is networking. they certainly don't need glasshouse levels of, say, humidity control. I guess power might be a question: compare a UPS-protected datacenter versus office where plugs are kicked... it’s a little hard to see the point here. yes, you can do HPC on NOW, and you can “serverize” the NOW with IPMI and cages. so the take-home is mainly that cube people need displays, and it’s cheap to attach a display to each node of the cluster? OK, sure – it’s not that it doesn’t make sense, but that it’s not news. and it’s slightly dubious in that it goes against thinclient/virtualization trends, and also assumes the workload is not network-sensitive. (I would claim that the latter is actually the _defining_ feature of HPC.)

this is almost a spread-the-pain kind of thing: take servers out of datacenters and spread them around the human-occupied spaces. noise is mostly a design issue, and at least in climates where buildings have to be heated, you get some small green benefit.

in other words, the main thing keeping servers in the machineroom is networking. they certainly don’t need glasshouse levels of, say, humidity control. I guess power might be a question: compare a UPS-protected datacenter versus office where plugs are kicked…

]]> By: Skip http://insidehpc.com/2010/09/15/intels-missing-middle-revives-attack-of-the-killer-micros/#comment-255729 Skip Wed, 15 Sep 2010 16:59:46 +0000 http://insidehpc.com/?p=13551#comment-255729 Isn't this GRID?? Isn’t this GRID??

]]>