nate wrote: > Svante Signell said: > >>Hi, > > >>I've always thought SCSI disks are faster than the IDE disks, but this >>does not seem to be the case (at least for device reads). Anyone have a >>good explanation, or am I missing something? > > > as another pointed out, your scsi disk is pretty old, keep in > mind that transfer rate isn't everything, access time is also > as important if not more so then transfer rate. many SCSI disks > have access times rated at sub 7ms (some are at 5ms). Most IDE > disks are in the 10+ms range still. One of the fastest IDE drives > is the Western Digital 8MB cache series(Special Edition) they have > 8.9ms access times..which is still higher then a 5400RPM IBM > 2.01GB ultrawide scsi disk I had back in 1996 ..(7.5ms). I have > 2 100GB Special ediiton drives in software raid1 connected to > a promise ata/100 controller, they are pretty fast. They don't > compare to my Ultra160 SCSI disk connected to a 29160N at work > though. > > also hdparm probably isn't the best tool to benchmark SCSI disks since > it is geared towards IDE disks.. > > bonnie++ can probably give more real-world tests since it operates > on the filesystem level and copies/creates/deletes files of many > sizes in several different ways. It also uses a large amount of data, > the tests you posted seem to indicate 64MB worth of data is all that > was measured, bonnie++ usually uses (system ram)*2 worth of disk > space to test. > > SCSI doesn't really start to shine though until you have multiple > devices on the bus. >
Thanks for the pointer to bonnie++. A few friends and I, who are sort of "advanced newbies" have just started looking at our system(s) performance. As indicated in my previous message, I discovered the default HD setup I had was not quite "optimum" and was seriously capping (limiting) my network file transfer speeds. Others in our group have reported similar observartions. We didn't know about bonnie++.... Looks like a very good tool for what we are looking at! Thanks Again!! Cheers, -Don Spoon- -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
