On Thu, Dec 24, 2009 at 8:18 PM, mike <xha...@gmail.com> wrote: > Like I said, you've never had to work in enterprise level areas. You need > a > LOT of I/O when 300 users are all hitting the same sql database, reading > and > writing to it all at once. A hard drive? The network would grind to a > halt > and no work would get done. > > Except that modern benchmarks don't show any appreciable performance improvement from RAID. Maybe 1-3%.
It used to be a huge difference when hard drives were slower and more expensive. And please don't use "enterprise" as if it were an example of good engineering. I work at the enterprise level and have for years and I see more stupid things done by big enterprises with big IT staffs than in most SMBs. So many people there want to use the "old, reliable" methods. Even when it no longer makes much sense. Remember when this first came up? It was mostly because organizations like Google and Amazon (cloud, EC2, etc.) can't afford to use RAID any more. Too expensive, too unreliable, too many failures, and not much benefit, even potentially. Enterprises can certainly afford to do their own benchmarks. Have your enterprise done one lately? In fact, if you want high availability and high performance, you are either massively redundant, like Google, or not even having hard drives in every machine. Too much work to replace a machine with a hard drive in it with little benefit. -- John DeCarlo, My Views Are My Own ************************************************************************* ** List info, subscription management, list rules, archives, privacy ** ** policy, calmness, a member map, and more at http://www.cguys.org/ ** *************************************************************************
