Comments on: “…and in the darkness bind them.” http://insidehpc.com/2008/02/06/and-in-the-darkness-bind-them/ 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: “…and in the darkness bind them.” http://insidehpc.com/2008/02/06/and-in-the-darkness-bind-them/#comment-36465 “…and in the darkness bind them.” Fri, 22 Feb 2008 16:15:18 +0000 http://insidehpc.com/2008/02/06/and-in-the-darkness-bind-them/#comment-36465 [...] MEDDESKTOP wrote an interesting post today onHere’s a quick excerpt…clusters of commodity computers, an approach that we postulate is akin to building a power plant from a collection of portable generators. [...] [...] MEDDESKTOP wrote an interesting post today onHere’s a quick excerpt…clusters of commodity computers, an approach that we postulate is akin to building a power plant from a collection of portable generators. [...]

]]> By: HPC is indeed everywhere | sun http://insidehpc.com/2008/02/06/and-in-the-darkness-bind-them/#comment-34333 HPC is indeed everywhere | sun Wed, 13 Feb 2008 17:21:29 +0000 http://insidehpc.com/2008/02/06/and-in-the-darkness-bind-them/#comment-34333 [...] and the IBM researchers are looking to apply HPC scale resources to the problems of hosting the web here.  This might be an interesting re-incarnation of the Big Friggin’ Web Tone Switch that [...] [...] and the IBM researchers are looking to apply HPC scale resources to the problems of hosting the web here.  This might be an interesting re-incarnation of the Big Friggin’ Web Tone Switch that [...]

]]> By: clusteradmin.net blog http://insidehpc.com/2008/02/06/and-in-the-darkness-bind-them/#comment-33476 clusteradmin.net blog Sun, 10 Feb 2008 18:32:04 +0000 http://insidehpc.com/2008/02/06/and-in-the-darkness-bind-them/#comment-33476 What is interesting to me in the original Register's article is how IBM people noticed, that programming for web using stateless tools such as http is in fact (implicitly) parallel programming - because each little request gets served independently. However, clusters _are_ designed to run many smaller jobs efficiently whereas large SMP machines are best with large applications. Of course they can often outperform a cluster for small jobs as well, but their high price usually convince people to use clusters for their smaller applications and buy SMPs for big stuff. -marek -- <a href="http://clusteradmin.net" rel="nofollow">clusteradmin.net :: blog about building and administering clusters</a> What is interesting to me in the original Register’s article is how IBM people noticed, that programming for web using stateless tools such as http is in fact (implicitly) parallel programming – because each little request gets served independently.
However, clusters _are_ designed to run many smaller jobs efficiently whereas large SMP machines are best with large applications. Of course they can often outperform a cluster for small jobs as well, but their high price usually convince people to use clusters for their smaller applications and buy SMPs for big stuff.

-marek

–
clusteradmin.net :: blog about building and administering clusters

]]>