DiskSpeedTest

Performance on Mac Pro RAID

It’s prob­a­bly a bit late in the game to be review­ing the per­for­mance of a four year old prod­uct, but one more set of bench­mark results in the world can’t hurt much, and may shed some light on more gen­er­al ques­tions about RAID per­for­mance in dif­fer­ent configurations.

Test­ing was per­formed on an 8‑core 2008 Mac Pro with the Apple Mac Pro RAID card and four 15,000RPM SAS dri­ves. The dri­ves were for­mat­ted in between tests and the tests were per­formed with no data on the dri­ves, so these results are nei­ther sci­en­tif­ic nor are they real-world results. But since the tests were per­formed on emp­ty dri­ves, we at least know that appli­ca­tion-spe­cif­ic fac­tors were not part of the results.

The tests were per­formed with Black­Mag­icDe­sign’s Disk Speed Test, which uses incom­press­ible data to test the speed of trans­fers with­out the aid of com­pres­sion algo­rithms that might arti­fi­cial­ly boost the per­for­mance numbers.

Though RAID 5 is often tout­ed as a bal­ance between safe­ty and per­for­mance, a vari­ety of sources around the Inter­net refer to the poor per­for­mance of RAID 5. One of the goals here was to deter­mine how much of this is hyper­bole and how much is truth.

Here are the raw numbers:

JBOD, sin­gle driveRAID 0, two drivesRAID 0, three drivesRAID 0, four drivesRAID 5, four drivesRAID 0+1, four drivesSin­gle SSD
write130MB/sec255MB/sec270MB/sec260MB/sec215MB/sec195MB/sec245MB/sec
read120MB/sec245MB/sec285MB/sec290MB/sec195MB/sec250MB/sec260MB/sec

The SSD speed test is not from the same sys­tem and is includ­ed only as a reference.

These results sug­gest that RAID 5’s rep­u­ta­tion for poor per­for­mance is some­what deserved — RAID 0 with only two dri­ves beats it by a sig­nif­i­cant mar­gin — but RAID 5 is still much faster than JBOD. A four-dri­ve RAID 0 is the fastest, but only by a very small mar­gin over a three-dri­ve RAID 0. (In this test the write per­for­mance was actu­al­ly slow­er with four dri­ves than with three, but even though the tests were ver­i­fied mul­ti­ple times the dif­fer­ence seems slight enough to con­sid­er this result a fluke.)

Sur­pris­ing­ly, RAID 0+1’s write per­for­mance was slow­er than any of the con­fig­u­ra­tions. While RAID 0+1’s read per­for­mance is respectable the write per­for­mance is the slow­est of all the test­ed con­fig­u­ra­tions but JBOD. Of course, RAID 5 and RAID 0+1 offer reli­a­bil­i­ty that RAID 0 does not, but except for high-avail­abil­i­ty sys­tems for which down­time is not an option, it seems that RAID 0 plus a good back­up and a spare dri­ve or two on hand would be an obvi­ous choice due to the per­for­mance benefit.

Again, these num­bers may be very dif­fer­ent with a dif­fer­ent RAID con­troller or with a soft­ware RAID, and not just faster or slow­er. Dif­fer­ent con­trollers may have total­ly dif­fer­ent pro­files or be opti­mized for par­tic­u­lar con­fig­u­ra­tions over others.

No con­fig­u­ra­tion comes close to the 553MB/sec Apple claims you can get from their hard­ware RAID card. It should­n’t be a sur­prise to see real-world bench­marks fall short of a man­u­fac­tur­er’s claims, but it real­ly would be nice to see esti­mates in the same ball­park as the best results that might actu­al­ly happen.

5 Replies to “Performance on Mac Pro RAID”

  1. Inter­est­ing scores…

    Inter­est­ing results. I am sur­prised that the two-dri­ve RAID0 match­es the per­for­mance of the SSD, but that sure makes the RAID the bet­ter price/performance option by far.

    You should note, of course, that your emp­ty dri­ves were work­ing on the front end of the dri­ve, which is typ­i­cal­ly twice as fast as the back end, and the SSDs don’t have that lim­i­ta­tion. Also, SSD access times are much faster for ran­dom access. (Most client oper­a­tions are gen­er­al­ly sequen­tial, which is why most bench­marks list sequen­tial R/W speeds.)

    Was your sin­gle dri­ve test done through the RAID con­troller? It seems low for such fast dri­ves. On Anandatech.com a sin­gle Veloci­Rap­tor 1TB SATA dri­ve gives sequen­tial read/write rates of 213 MB/s on the out­er tracks and aver­age 173MB/s over­all. Is that because they are new­er than your 15k SAS dri­ves? Or is it sim­ply a dif­fer­ent test?

    Dad

    1. Price/performance

      I’m not sure that this shows that the RAID makes for the best price/performance option «by far». OWC sells a 480GB SSD that’s a bit faster than the one in my Mac­Book Air for under $600. A decent RAID con­troller that sup­ports SAS will cost $250 (you could get the same one I have for $600 — but why would any­one do that?) and 300GB 15,000 RPM SAS dri­ves will run $175 each, so the two-dri­ve RAID will cost about the same as the SSD. True that the two-dri­ve RAID will have 120GB more stor­age, but as you point out the SSD will have faster access times. The SSD will also con­sume less pow­er and be qui­eter. I’d say at best RAID edges out the SSD at the stor­age sizes I tested.

      Go up or down the capac­i­ty scale from there and the price/performance ratios change dra­mat­i­cal­ly. Drop down to a 240GB SSD and the SSD is less expen­sive than the RAID con­troller alone. Upgrad­ing to a pair of 450GB 15,000RPM SAS dri­ves for a total of 900GB only brings the RAID price up to about $700, while a 960GB SSD is $1270. No ques­tion that when you include capac­i­ty, RAID will scale much more eco­nom­i­cal­ly. Even a pair of 600GB 15,000RPM SAS dri­ves plus con­troller comes up cheap­er than the 960GB SSD.

      Con­sid­er also, how­ev­er, that I have a (rel­a­tive­ly) poor-per­form­ing SSD in the 2011 Mac­Book Air. OWC’s $730 480GB SATA SSD will blow mine away in read through­put and per­form mod­er­ate­ly bet­ter in write throughput.

      I was­n’t able to find the same results on anandtech.com that you did, but I’d be inter­est­ed to see whether the results were from mov­ing «typ­i­cal» data or incom­press­ible data. In any case, it prob­a­bly does­n’t say very much to com­pare results arrived at by dif­fer­ent sets of tests.

      Though it’s only 10,000RPM, the Veloci­rap­tor is basi­cal­ly a 2.5″ dri­ve in a 3.5″ heatsink, which means the plat­ters them­selves have to be small­er. With 1TB on a 2.5″ dri­ve, the plat­ter den­si­ty must be much high­er than 300GB on a 3.5″ drive.

      The sin­gle dri­ve test was done through the RAID con­troller. I don’t have anoth­er con­troller to swap out. I too was sur­prised by it, but I checked all four dri­ves and they all per­formed the same. (Part of the point of this test­ing was to make sure that I did­n’t have one mar­gin­al dri­ve drag­ging down the per­for­mance of the whole array.)

  2. Most PC moth­er­boards have

    Most PC moth­er­boards have RAID con­trollers built-in, negat­ing the cost of the add-in card. If a sin­gle Veloci­Rap­tor is almost as fast as your SSD, I would guess a pair of stan­dard 7200rpm SATA Bar­racu­d­as in RAID0 would match it, at a total cost of around $200 for two 1TB dri­ves. This is all guess­work on my part, but that seems pret­ty like­ly. Or Ama­zon is sell­ing the 1TB Veloci­Rap­tor for $215 each.

    Here’s the link to the Anandtech review. The sequen­tial tests are toward the bot­tom. http://www.account.anandtech.com/show/5729/western-digital-velociraptor-1tb-wd1000dhtz-review/2

    For most users Anand recom­mands a small 128GB SSD for the boot dri­ve and RAID­ed 7200rpm dri­ves for stor­age and reserves the Veloci­Rap­tor for work sta­tions or game machines.

    With a spe­cif­ic appli­ca­tion like you have, I think stan­dard bench­marks are prob­a­bly only curi­ousi­ties. How Pho­to­shop caching, RAM size, OS, etc inter­act prob­a­bly makes as much dif­fer­ence as any­thing. Your image files are sig­nif­i­cant­ly larg­er than most peo­ple can even comprehend.

    Dad

     

     

    1. Most on-board RAID

      Most on-board RAID con­trollers require the main CPU to do a lot of the work of con­trol­ling the RAID, mak­ing them some­what bet­ter than soft­ware RAID but not as good as a sep­a­rate con­troller. Though it depends great­ly on which moth­er­board and which chipset you get.

      I bought this sys­tem four years ago. If I had it to do over again, I’d prob­a­bly do some­thing sim­i­lar to what you’ve described. I might use a larg­er SSD to acco­mo­date enor­mous Pho­to­shop swap files, or I might get two SSDs, most­ly because admit­ting that one has one’s swap file on one’s sys­tem vol­ume is a good way to not get one’s ques­tions answered in sup­port forums, even if the per­for­mance ben­e­fits of hav­ing swap and scratch on sep­a­rate devices is mar­gin­al, as I’ve shown.

      Then I’d sup­ple­ment the sys­tem with a big dumb slow hard dri­ve (which of course would­n’t be that slow — I’m still talk­ing SATA, not USB or FireWire) so that I’d have plen­ty of space for stor­age on my sys­tem with the fast boot and scratch drives.

      Recon­fig­ur­ing the sys­tem as I have, with three dri­ves as RAID 0, a fourth sit­ting spare, and Time Machine mak­ing hourly back­ups, gives me a bit more speed than I had, almost two hun­dred more gigs of space, and it does­n’t cost me any­thing today.

      There are some lim­i­ta­tions on my cur­rent hard­ware which will shape any upgrade plans. The Mac Pro RAID card is not one that any­one rec­om­mends unless you’re buy­ing it with a new sys­tem and are too lazy to buy a RAID con­troller sep­a­rate­ly. The 2008 mod­el of the Mac Pro RAID card specif­i­cal­ly is one that some peo­ple rec­om­mend just be chucked in the trash rather than deal with the lim­i­ta­tions and lack­lus­ter performance.

      I’ll need to check and see whether I can mix SAS and SATA dri­ves on this con­troller. I’m not look­ing to build a RAID with a mix­ture of dri­ves, but I might want to buy a nice 2TB or big­ger dri­ve any put it in the fourth bay to give me more space and to let my fourth «spare» SAS dri­ve spin down and rest for the day it has to be pressed into service.

      Although even if I can’t mix SAS and SATA on the same card, SAS dri­ves aren’t that expen­sive if I don’t need a 15,000RPM mod­el. Not as cheap as SATA dri­ves, but I prob­a­bly don’t need to rule out SAS for that fourth bay, just rule out 15K.

      I know that I can­not put any kind of SSD on this con­troller, so if I do add SSD to this machine, it will be a PCIe card like the one that OWC sells. It won’t need to be very big, as you say, but I’ll have to make the price ver­sus capac­i­ty choice based on the prices when I do that. I’ve seen the prices drop twice in four months already so there’s real­ly no telling whether it will seem like a good idea to go with 240GB or 120GB. Or 960GB. Yes, I’m kid­ding, but it is tempt­ing to replace those 15,000RPM dri­ves if for no rea­son oth­er than the con­stant noise they make. Anoth­er draw­back of the 2008 mod­el of my con­troller is that it does not sup­port putting dri­ves to sleep when they aren’t in use. As a result there’s no such thing as sleep mode on my Mac Pro.

      Like I said, if I were buy­ing the machine today, there are things I’d do dif­fer­ent­ly. I just spec’d out an iMac on the Apple site: 3.4GHz quad-core i7 (which would blow away my dual quad-core 2.8GHz Xeons), 16GB of RAM (admit­ted­ly less than the 32 I have now), 256GB SSD plus 2TB 7200 RPM dri­ve, and the upgrad­ed video con­troller for the 27″ screen (again, less than the 30″ dis­play I don’t have any longer.) I basi­cal­ly spec’d out the most expen­sive iMac I could and it came to about three-eighths of what I spent on the Mac Pro. If I sourced my mem­o­ry from OWC I could get it to 32GB and only pay a total of about $20 more.

      But that was­n’t avail­able four years ago. This pup­py still holds it own pret­ty well and I sus­pect it will con­tin­ue to for some time yet.

  3. Adden­dum

    My RAID con­troller’s bat­tery is in a con­di­tion­ing cycle and as a result my RAID cache is dis­abled, so I decid­ed it might be infor­ma­tive to try out the speed test with­out the cache. I got 160MB/sec writes and 285MB/sec reads from the test with my cur­rent 3‑disk RAID0 con­fig­u­ra­tion. On the remain­ing JBOD disk I got 105MB/sec writes and 125MB/sec reads. Both sets of results tell me (not sur­pris­ing­ly) that the cache speeds up writes but does noth­ing for read performance.

Leave a Reply