Re: Disk Sizes

Bill, One approach is to balance the total I/O demand of your server(s) with the total I/O bandwidth of your disks and I/O channels. I've never seen or heard of any non-mainframe processor running any DBMS that will ask for more than 10 MB per second (sequential) or 800 I/O per second (random). I've posted here a couple of times but no one has ever responded that they've seen higher rates. Given the "speeds and feeds" for current 3.5" disks, that works out to approximately one disk per processor (for sequential I/O) and ten disks per processor (for random I/O). I/O paths are currently limited at 16MB/sec (SCSI FWD), 33MB/sec (SCSI UWD) and 60-70MB/sec (FCAL). For example, it's highly unlike that XPS on a 4-node system with 4 processors each (16 total) would ever need more that 160 disks hung off 6 SCSI UWD channels. Even the 5.25" 5400rpm disks can be appropriate for DW environments because their sequential transfer rates are still >>15MB/second. Their random rates are lower because of larger seek and rotational delay times. Doug Bill Weaver wrote in message <74h9g0$a1p$1@news.xmission.com>... > >I have mostly 4 Gig drives in my system with a few 2 Gig drives. I'm looking to upgrade the 2 >Gig drives to 4. My vendor suggested I get 9 Gig drives instead (price per Gig actually is >cheaper for 9 than 4 now since 4's aren't used as much and they aren't making as many). I seem >to remember a discussion on this list a while back that indicated that it would be better to >have the 4 Gig drives over the larger drives but for the life of me I can't remember the >reasoning. Can someone enlighten me? Is it simply that if you have more smaller drives then >you can spread the work out more among the drives or are there other reasons as well? Thanks >for your help! >-------------------------------------------------------- >Name: Bill Weaver >E-mail: Bill Weaver <billw@fscorp.com> >Date: 12/07/98 >Time: 13:43:26 > >This message was sent by Z-Mail Pro - from NetManage >NetManage - delivers Standards Based IntraNet Solutions >-------------------------------------------------------- >