I've got 6x WD2003FYYS 2TB drives connected directly to an Adaptec 51245 controller. I have been experimenting with different configurations to determine which is the best for storing my Blu-Ray and DVD ISO files, along with family videos, pictures, music etc. So ideally I'd like RAID 6 as it gives some redundancy - I can't back up all the ISO's (although less of an issue as I have originals), but will back up the other files - although RAID 6 should give me a bit of extra comfort about drive failure too. Be that as it may, I've run ATTO benchmarks on the RAID 5 and 6 configurations with an odd outcome. On average, they both provide around 575MB/s write speeds (I would have thought RAID 6 would be a bit slower) but the read speed is the interesting bit. RAID 5 gets around 1680MB/s but RAID 6 maxes out at about 600MB/s. Any thoughts on why this is the case? I probably don't need maximum read speed, so perhaps I should stick with RAID 6, but for now I'm running RAID 5 with a hot spare. For my purposes (mainly streaming) faster read speed is the better option. However, I'm a bit confused - I would have expected the two read speeds to be similar.
RAID 5 vs 6 speed issue - any ideas?
- Thread starter gaking
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Six spinning drives in Raid-5 should never get more than about 5x the read speed of a single drive. Any more than that means that you are not reading the speed of your drives - you are reading the speed of the cache on your controller. The speeds you are seeing on the "raid-6" configuration are rational. The speed you are reporting for the Raid5 is not.
Things that look to good to be true usually are.
Things that look to good to be true usually are.
I would also add that any speed difference between R5 & R6 will not be realized over a home network, so its essentially a moot point. I would pick R6 for added redundancy.
I'm not clear what that has to do with this issue? I realise that the speed of a Blu-Ray may be way less than I'm capable of generating, but I'm more interested in the differences between RAID 5 and 6 than the throughput of the Blu-Ray since either configuration I've tested is way in excess of the Blu-Ray speed anyway.54 Mbps is the maximum bit-rate for a BD-ROM. That is less than 7MB/s.
Last edited:
I would tend to agree with you - it's just odd that running identical ATTO benchmarks between the two configurations gives such wildly varying results, i.e. why would one show the cached results and the other not? Perhaps I should turn off the read cache for both tests to just get a better comparison.Six spinning drives in Raid-5 should never get more than about 5x the read speed of a single drive. Any more than that means that you are not reading the speed of your drives - you are reading the speed of the cache on your controller. The speeds you are seeing on the "raid-6" configuration are rational. The speed you are reporting for the Raid5 is not.
Things that look to good to be true usually are.
Thanks, I was starting to go down the same route, but was then toying with the fact that the hot spare would be powered down for most of the time and provide a power saving whilst in R6 it will be constantly spinning. However, there is obviously the reliability issue with another failure during a rebuild of R5 vs R6 dual parity redundancy - so I'm still not 100% sure on which way to go. The added "performance" of R5 looked like a desirable factor, but as has been said it's unlikely to be a) true or b) of much use in a home environment. Still, it would be nice if it were believable. For what it's worth, I previously had WD2002FYPS drives in R5 but I RMA'd them as they had some compatibility issues with the Adaptec card (one of them being only able to connect at 1.5gbps); ATTO reflected about 1400MB/s in those tests....I would also add that any speed difference between R5 & R6 will not be realized over a home network, so its essentially a moot point. I would pick R6 for added redundancy.
You wrote in your OP: "For my purposes (mainly streaming) faster read speed is the better option. However, I'm a bit confused..."I'm not clear what that has to do with this issue? I realise that the speed of a Blu-Ray may be way less than I'm capable of generating, but I'm more interested in the differences between RAID 5 and 6 than the throughput of the Blu-Ray since either configuration I've tested is way in excess of the Blu-Ray speed anyway.
You certainly do seem confused. Obviously, faster read speed is not the better option -- it is irrelevant. That is what the 54 Mbps bit-rate has to do with the issue. I thought that was obvious, but apparently not.
ATTO is easily confused by controller cache. This is a long-standing and well documented issue with that benchmark. You should always turn off the read cache when using ATTO (and - for that matter - most other disk benchmarks).I would tend to agree with you - it's just odd that running identical ATTO benchmarks between the two configurations gives such wildly varying results, i.e. why would one show the cached results and the other not? Perhaps I should turn off the read cache for both tests to just get a better comparison.
I'm afraid you are misinterpreting or misreading what I said. What I said was that I was confused / unsure why RAID 5 and RAID 6 read speeds would be so dissimilar with the same benchmark tool under the same test conditions other than the RAID level. Write speed for level 6 should generally be slower (additional parity to be recorded), read should be about the same between the two levels. And whilst I can understand that the read cache can / will cause overstated results, I still find it odd that the cache is read in the test for RAID 5 but the disc array performance in RAID 6. I would have expected the same tool to analyse the same information in the same manner irrespective of the underlying RAID level of the controller. So yes, whilst the test was perhaps flawed in that I should have disabled the read cache first (my bad), it should have given similiarly overstated speeds in both sets of circumstances, but it didn't - that is what I am trying to fathom.You wrote in your OP: "For my purposes (mainly streaming) faster read speed is the better option. However, I'm a bit confused..."
You certainly do seem confused. Obviously, faster read speed is not the better option -- it is irrelevant. That is what the 54 Mbps bit-rate has to do with the issue. I thought that was obvious, but apparently not.
I am not confused at all about the fact that if I am streaming content, the faster the read speed the better since it allows multiple simultaneous streams. Write speed is what is irrelevant as nothing is being written to disc. However, as I said - the controller is nevertheless capable of far more than one stream of Blu-Ray data so if you are only considering it from that aspect then fine, your point is valid and I understand it. However, that was not the point of my question, so contrary to what you are saying I'm afraid that you are the one who is confused, not me - the question was about RAID 5 vs 6 speed read differentials, not whether the throughput was adequate for a single Blu-Ray stream. Anyway, I don't really want to debate that point further - all I really would like to understand is why ATTO would give such different results or why the cache seems to be relevant in one set of tests but not in the other when only the RAID level changes. For example, is it saying that perhaps RAID 6 is inefficiently cached on the Adaptec controller for some reason? It may be a bit academic, but I would like to understand why, that is all. I still prefer level 6 as it has extra redundancy with more than sufficient performance.