Which one of these is best for a RAID5 array?
- Thread starter voodooFX
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Even with a UPS, almost all RAID cards will perform dramatically better with a BBU since you can then safely enable write caching. So for your RAID5 scenario, I'd only consider cards with battery or capacitor backed caches.
First off, @voodooFX I love your avatar. I still remember my boards fondly.
You may want to look at something like an IBM M5014. 256MB cache only, but you can get them fairly inexpensively, often with BBUs. I think the last few I sold were in the $140 range. Granted, those will have older batteries so that is something to take into account as well.
You may want to look at something like an IBM M5014. 256MB cache only, but you can get them fairly inexpensively, often with BBUs. I think the last few I sold were in the $140 range. Granted, those will have older batteries so that is something to take into account as well.
Thanks for the suggestions
I'm building my storage server and since I already have all the listed cards I was hoping there is one acceptable as RAID5 controller..
I'm confused: there is a performance difference between write-back-cache enabled (after the usual warning) on a card without the BBU and a card with the BBU?
@Patrick thanks, I have almost all the voodoo collection, V1, Rush, V2 (V2SLI), Banshee, V3 (2000-3000-3500TV), V4 4500 and V5 5500
I'm building my storage server and since I already have all the listed cards I was hoping there is one acceptable as RAID5 controller..
I'm confused: there is a performance difference between write-back-cache enabled (after the usual warning) on a card without the BBU and a card with the BBU?
@Patrick thanks, I have almost all the voodoo collection, V1, Rush, V2 (V2SLI), Banshee, V3 (2000-3000-3500TV), V4 4500 and V5 5500
Heh heh, I had a few Voodoo cards in the past too:
Pair of Creative VoodooII cards overclocked to 110MHz and SLI'ed (passing through a single-head Matrox G400)
Voodoo 3-3000, I loved that card
Finally, a V5-5500. HOT. Hotter than hot. Totally beat the hell out of Unreal engined games, though.
Pair of Creative VoodooII cards overclocked to 110MHz and SLI'ed (passing through a single-head Matrox G400)
Voodoo 3-3000, I loved that card
Finally, a V5-5500. HOT. Hotter than hot. Totally beat the hell out of Unreal engined games, though.
Why not PERC 5/i? Works fine for me with 4x RE4 1TB with good performance. They're like $10-25 all day long on ebay.
and the winner is.. Adaptec 5405 | 256MB DDR2 | w/ BBU
It gives me about 350+ MB/s (seq.) in read/write which is more than I expected.. and the good thing is that I have two of them so if the first decide to die I have a substitute
FMA1394: I was looking for a solution with the controllers I already have, but thank for the suggestion
It gives me about 350+ MB/s (seq.) in read/write which is more than I expected.. and the good thing is that I have two of them so if the first decide to die I have a substitute
FMA1394: I was looking for a solution with the controllers I already have, but thank for the suggestion
Just curious... what server do you have them in? I would expect 4 drives to scale a bit more linearly...and the winner is.. Adaptec 5405 | 256MB DDR2 | w/ BBU
It gives me about 350+ MB/s (seq.) in read/write which is more than I expected.. and the good thing is that I have two of them so if the first decide to die I have a substitute
FMA1394: I was looking for a solution with the controllers I already have, but thank for the suggestion
4 drive raid 5 was getting me more like 5-600MB/s on RE3s
I thought I was getting closer to 150 per drive...
Looks like I remember that array poorly...
And I don't have numbers for my 7200.14 array though throwing 8GB across it took no time at all.
Your results look very strange, I mean they seems to be very variable and heavily influenced by the cache.
What controller were you using?
Do you remember what was your "real" sequential speed, moving for example 5+ GB of big files (iso, mkv..) from and to the array?
this is what I'm getting with CDM.
This is with 4x WD RE4 500GB.
In the final build I will use RE4 1TB
What controller were you using?
Do you remember what was your "real" sequential speed, moving for example 5+ GB of big files (iso, mkv..) from and to the array?
this is what I'm getting with CDM.
This is with 4x WD RE4 500GB.
In the final build I will use RE4 1TB
HP SA (Smart Array) P420 4x 500GB RE3s not sure if a drive was bad or if it was just the early firmware I had on it at the time.
The cache is super fast on those... but on my array of 4x 7200.14s I would see windows transfers in the 700MB/s range for 15GB files.... however I have no doubt windows was doing some heavy caching on me... side effects of having 64GB installed.
I will get more benchmarks done on that controller when I rebuild the array. With and without SSD caching.
You seem to be getting quite solid of numbers.
PMC builds solid hardware. HP SA use the same controllers and now PMC has the license to use SA firmware and drivers.
I would not recommend using SSDs on Adaptec controllers for another few months... the SA devs are hard at work on the firmware and drivers.
I think you have chosen wisely.
The cache is super fast on those... but on my array of 4x 7200.14s I would see windows transfers in the 700MB/s range for 15GB files.... however I have no doubt windows was doing some heavy caching on me... side effects of having 64GB installed.
I will get more benchmarks done on that controller when I rebuild the array. With and without SSD caching.
You seem to be getting quite solid of numbers.
PMC builds solid hardware. HP SA use the same controllers and now PMC has the license to use SA firmware and drivers.
I would not recommend using SSDs on Adaptec controllers for another few months... the SA devs are hard at work on the firmware and drivers.
I think you have chosen wisely.