Just as an FYI on the PCI bus, we found out from some testing that a 64-bit, 66MHz PCI bus can be the bottleneck while driving two LTO Ultrium tape devices, so the PCI bus could definitely be a bottleneck for fast hard disks (I don't know about ATA100/133 IDE, but with that many disks, probably). The next step up would be PCI-X which (I believe) is currently running at 133MHz and provides burst transfer rates at up to 1GB/s (GigaByte/sec) . Regards, Andy. -----Original Message----- From: Alvin Oga [mailto:aoga@Maggie.Linux-Consulting.com] Sent: Wednesday, August 21, 2002 5:07 PM To: Maurice Hilarius Cc: linux-raid@vger.kernel.org Subject: Re: large devices (3ware) - thanx for the info hi ya maurice yes... thankx for the benchmark numbers ... good that the pci buss is the bottle neck for the disk io transfers ( which would than have one wonder if the 64bit pic buss ( or the so-called 533FSB is any better and yes... the file sizes does affect overall transfer speeds etc thanx alvin and nope... we only use 1U chassis... so we're stuck with one pci card using the normal off-the-shelf atx stuff On Wed, 21 Aug 2002, Maurice Hilarius wrote: > With regards to your message at 01:53 PM 8/21/02, Alvin Oga. Where you stated: > >3ware has a 8-port ide card ... for 8 master ide disks but its > >still on one pci buss ( i think thats the bottle neck ) > And a 12 port card too. > > >if you have 12, 26 ide disks... you'd need a 2nd pci slot...but again > >on the one pci buss... > We do this all the time. 2 of the cards. One array. > > >i think it'd be better to put one ide controller ( like the 3-ware ) > >into one motherboard... and use a different server for the next 8 disks > Don't need it. > Bottleneck is that the card is 64 bit, but only 33MHz. > It's throughput maxes around 160MB/sec. > This is a limitation of the PCI at 64/33 > > > >-- am looking for benchmarks of 8-24-48 hard disks systems > > Here is for 2 x 8 port cards, 16 x Maxtor 160GB. > RAID5, on a Tyan S2460 AthlonMP board with a single Duron CPU. > Cards in JBOD mode, using Linux software RAID. > > Version 1.02a ------Sequential Output------ --Sequential Input- --Random- > -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- > --Seeks-- > Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP > raiddawg 4G 17752 89 78709 44 55648 32 20109 98 162752 67 > 224.3 1 > ------Sequential Create------ --------Random > Create-------- > -Create-- --Read--- -Delete-- -Create-- --Read--- > -Delete-- > files /sec %CP /sec %CP /sec %CP /sec %CP /sec > %CP /sec %CP > 16 1240 99 +++++ +++ +++++ +++ 1284 99 +++++ > +++ 3581 99 > raiddawg,4G,17752,89,78709,44,55648,32,20109,98,162752,67,224.3,1,16,124 0,99,+++++,+++,+++++,+++,1284,99,+++++,+++,3581,99 > > With a faster CPU the writes go up higher, typically up to 90MB/sec. > Reads still max around 160MB/sec. > With a single card and 8 drives we get the same type of numbers. > With 2 of the 12 port cards and 24 disks we still get the same numbers. > > Note these tests use 4GB file size. > If you use some smaller size you can get higher figures, but is essentially > bullshit, as these are then influenced by Linux cacheing, drive caches, etc.. > > > With our best regards, > > Maurice W. Hilarius Telephone: 01-780-456-9771 > Hard Data Ltd. FAX: 01-780-456-9772 > 11060 - 166 Avenue mailto:maurice@harddata.com > Edmonton, AB, Canada http://www.harddata.com/ > T5X 1Y3 > > 2.3TB RAID5 NAS server - dual AthlonMP CPU, Linux, $10,995 CAD / $6850 USD > > - > To unsubscribe from this list: send the line "unsubscribe linux-raid" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > - To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html - To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
