I have no idea as to what the tier1 vendors say as I have only worked within the storage business.. the figures I quoted are based on the last time I consulted on this are would been provided by IBM / Seagate as these are the only two scsi vendors we use. If you really want to dig, then ask Seagate, they are respected in both camps and will openly justify the technology and price difference. They produce extremely in-depth docs on the testing methods and assumptions. In terms of reset I am not sure what you mean... we and all raid manufacturers will reset a scsi bus on scsi timeouts.. this is normal practice and simple to achieve. It is not achievable on sata.. I have not used pata much, but I do not recall a reset line that we could trigger from firmware level. RAID in isolation does not increase the i/o load as we all know... but the reality is that raid applications do. Non of us can refuse the cost effective nature of sata drives, this means we can often use raid in places where we could not afford or justify scsi. Add multiple users and the stress on the drives increase dramatically. If you want a real life situation... one of our scsi designs is used around the world and has probably 10m+ users (many systems).. in some cases these have been running for 4 / 5 years and therefore we have to look at drive replacement. For a trial we used sata to obviously see if we could save costs or offer an intermediate solution. We could not keep a single system going for more than 14 days. The load varied between 10-250 users at any one time.. we tried Maxtor and IBM. There was also a 40% occurrence of fatal state errors.. this was simple the rate that the drives were failing meant it was likely to fail whilst in rebuild state and obviously die. Take the sata box and stick it in many applications and it will last you to your dying day. You may be right that there has been ata and scsi drive manufactured with the same components excluding the interface.... but the last time I saw this was a bearing shortage in 95... I don't know of any manufactures today that even hint at this. But I could well be wrong.. The discussion could probably go on forever, but the point is that we are not stupid... sata solutions are probably 30% of the cost of the scsi..... there is a difference and we know it. the important thing is accepting the difference and using the right technology for the right application. -----Original Message----- From: Mark Hahn [mailto:hahn@xxxxxxxxxxxxxxxxxxx] Sent: 20 November 2004 21:58 To: Mark Klarzynski Subject: RE: Good news / bad news - The joys of RAID > SATA / IDE drives have an MTBF similar to that of SCSI / Fibre. But this > is based upon their expected use... i.e. SCSI used to be [power on hours > = 24hr] [use = 8 hours].. whilst SATA used to be [power on = 8 hours] > and [use = 20 mins]. can you cite a source for these numbers? the vendors I talk to (tier1 system vendors, not disk vendors) usually state 24x7 100% duty cycles for scsi/fc, and 100% poweron, 20% duty cycles for PATA/SATA. - To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
