Can SATA breach the chasm?
Posted in Storage Interconnects & RAID, Advisor - Tom by Tom TreadwayQuestion to the Storage Advisors, from anonymous: Do you think the stability and reliability of SATA drives will improve? We have not had a lot of luck with them in production and the overall opinion is that ATA is a desktop technology vs SCSI which is designed for the server end.
As you already know, SATA, ATA, SAS and SCSI are all interface specs and of course have nothing to do with the reliability of the drive itself. But as you’ve correctly pointed out, ATA drives have historically demonstrated a much lower reliability than SCSI. This lower reliability is caused by lower quality motors, heads, media, etc.
Reliability can be judged with two commonly used metrics: MTBF (Mean Time Before Failure) and BER (Bit Error Rate). MTBF is measured in hours and is the average time for a drive to fail. Some folks also like to use the metric AFR (Annual Failure Rate) to indicate the percentage of drives that will fail in a year. Of course that’s just the inverse of MTBF, so no additional information is conveyed. BER is the average number of bits that have to be read to before finding an uncorrectable bit. This metric is not so important in optimal redundant arrays like RAID-5, but it’s critical when trying to rebuild a degraded array. As drives have increased in size the BER has become a much bigger factor than MTBF in determining array failure rates.
A more detailed review of the equations can be found in my post on RAID Reliability.
Actual measured MTBF and BER for SCSI and ATA can be found in my post on Reliability Calculations.
So how did it come to be that ATA used “lesser” parts while SCSI used higher-end parts? It really started from the fact that ATA drives were typically found in workstations and home machines where low cost is critical, while SCSI drives were used in servers where performance and reliability were critical. You can therefore expect that as SATA replaces ATA, the MTBF and BER will remain the same. Likewise, as SAS replaces SCSI, the MTBF and BER will remain the same.
But SAS/SATA has one big difference over SCSI/ATA: SAS and SATA can coexist in the same storage domain. In other words, a single SAS controller can support both drive types – simultaneously. So this has created the opportunity for companies like Seagate to produce “tweener” drives such as the NL35, a nearline SATA drive with an MTBF closer to SAS than typical SATA drives. You can expect this blurring of entry-level and high-end to continue.
Lastly, I would be remiss in not pointing out how RAID-6 can be used to drastically increase the reliability of arrays based on SATA, as described in my post on RAID-6 Configurations.
TT