On 20 Jun 2006 at 13:07, Dan Langille wrote: > I just heard from a noted developer who has worked on RAID a lot. The > rest of this post is more or less his words and from his point of > view, not mine (ie. if I say I, assume he is speaking). > > For SCSI, you are better off having all the drives the same. He has > no data on SATA, but the star-topology nature of SATA may not be > prone to the same bus-tpology issues that SCSI has. > > With SCSI, each vendor has a slightly different interpretation of the > spec, though usually a legal one. The differences in timing, > reconnection policies, etc, by different brands on the same bus can > cause hic-ups under very high load. The controller might see a > spurious timeout on a device, for example, when the drive is actually > OK. For RAID controllers that aggressively defend against problems, > this can result in bogus failure reports. > > With SATA, since each drive essentially is alone on its bus, that > might not be an issue. But then again, it might be an issue for the > controller chip that has to talk to all of the drives. I don't have > much data on that, though. > > hth. In addition: ### If you're worried about these kinds of problems, then you should already be carefully selecting and qualifying your hardware before putting it into production. And, you should be doing regular backups ### The following is for those that do not know otherwise: RAID is not a backup. RAID lets the system survive a disk failure. RAID is not a backup. RAID without backup is ill-advised. -- Dan Langille : Software Developer looking for work my resume: http://www.freebsddiary.org/dan_langille.php