09-11-2022, 09:30 AM
Hi,
Let me first say I’m hugely appreciative of the YouTube channel :-)
Query: I have become confused about whether to consider UREs, and risk of RAID rebuild failure, when planning a home NAS build. I'm sure you're aware of the maths on chance of encountering a URE during a rebuild circulating on various forums, leading to calculators like this:
http://magj.github.io/raid-failure/
Considering a WD Red Plus (1 in 10^14) URE rate, and a 4 x 4TB RAID 5, this suggest a rebuild may have less than 50% chance of succeeding should a drive fail.
Where as using IronWolf (1 in 10^15) URE implies the equivalent NAS has >90% chance of rebuilding following a disk failure.
This is a dramatic difference in my view since one is tolerable and the other is not. This also implies that large (12+ TB) disks in something like a DS920 under RAID 5 are not a good idea. This can also guide one towards 8 (RAID 6) bays unit but obviously at higher cost.
I’d love to know if you have an opinion on using these calculation to determine a NAS build. I appreciate RAID 6 will always be more reliable than RAID 5, and obviously lower URE chance is always a good thing, but I’m concerned that these seemingly objective probabilities are potentially way off (these are URE limits afterall so they may be much lower I suppose in reality) and about letting them unduly influence a build decision. Ie give an impression that increased cost (eg from 4 x 8TB to 8 x 4TB unit) is buying more reliability than it actually is. As I say, these calcs dont seem to stack up with what people are doing in reality. Do you think the URE risk should be ignored?
Also, did you every cover this discussion - that seems to go back and forth in forums - in one of your (excellent) videos?
Many thanks,
Paul
Let me first say I’m hugely appreciative of the YouTube channel :-)
Query: I have become confused about whether to consider UREs, and risk of RAID rebuild failure, when planning a home NAS build. I'm sure you're aware of the maths on chance of encountering a URE during a rebuild circulating on various forums, leading to calculators like this:
http://magj.github.io/raid-failure/
Considering a WD Red Plus (1 in 10^14) URE rate, and a 4 x 4TB RAID 5, this suggest a rebuild may have less than 50% chance of succeeding should a drive fail.
Where as using IronWolf (1 in 10^15) URE implies the equivalent NAS has >90% chance of rebuilding following a disk failure.
This is a dramatic difference in my view since one is tolerable and the other is not. This also implies that large (12+ TB) disks in something like a DS920 under RAID 5 are not a good idea. This can also guide one towards 8 (RAID 6) bays unit but obviously at higher cost.
I’d love to know if you have an opinion on using these calculation to determine a NAS build. I appreciate RAID 6 will always be more reliable than RAID 5, and obviously lower URE chance is always a good thing, but I’m concerned that these seemingly objective probabilities are potentially way off (these are URE limits afterall so they may be much lower I suppose in reality) and about letting them unduly influence a build decision. Ie give an impression that increased cost (eg from 4 x 8TB to 8 x 4TB unit) is buying more reliability than it actually is. As I say, these calcs dont seem to stack up with what people are doing in reality. Do you think the URE risk should be ignored?
Also, did you every cover this discussion - that seems to go back and forth in forums - in one of your (excellent) videos?
Many thanks,
Paul
