Tech Stuff

Did some benchmarking on the UFies.org box (see the system specs here) to see if the nice 18G Seagate Cheetah drive, with an 8meg cache, compares to 4 Maxtor 40G drives in a RAID5 configuration, each with a pretty decent speed itself? Does SCSI win? Whose drive reigns supreme?


Sorry, slipped into Iron Chef mode for a moment there.


Running hdparm tests on the SCSI drive (/dev/sda) and an individula (non-RAID) IDE drive (/dev/hde) gave me the following results for three separate runs:


Run #1


/dev/sda:
Timing buffer-cache reads: 128 MB in 0.79 seconds =162.03 MB/sec
Timing buffered disk reads: 64 MB in 2.10 seconds = 30.48 MB/sec

/dev/hde:
Timing buffer-cache reads: 128 MB in 0.69 seconds =185.51 MB/sec
Timing buffered disk reads: 64 MB in 1.74 seconds = 36.78 MB/sec


Run #2

/dev/sda:
Timing buffer-cache reads: 128 MB in 0.56 seconds =228.57 MB/sec
Timing buffered disk reads: 64 MB in 2.48 seconds = 25.81 MB/sec

/dev/hde:
Timing buffer-cache reads: 128 MB in 0.55 seconds =232.73 MB/sec
Timing buffered disk reads: 64 MB in 1.60 seconds = 40.00 MB/sec


Run #3

/dev/sda:
Timing buffer-cache reads: 128 MB in 0.57 seconds =224.56 MB/sec
Timing buffered disk reads: 64 MB in 2.28 seconds = 28.07 MB/sec

/dev/hde:
Timing buffer-cache reads: 128 MB in 0.57 seconds =224.56 MB/sec
Timing buffered disk reads: 64 MB in 1.60 seconds = 40.00 MB/sec



After this I ran bonnie on the filesystem to test it. Bonnie not only takes into account the speed of data being moved around the disk, but also of the amount of CPU that it takes to move it. A page with how to interpret the bonnie results is here.


First for the RAID5 setup (/var) the command line was:


bonnie -d /var/tmp -s 1g -n master -r 512MB -u nobody:nogroup -f


And the results:




Version 1.02c ——Sequential Output—— –Sequential Input- –Random-
-Per Chr- –Block– -Rewrite- -Per Chr- –Block– –Seeks–
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
master 1G 22414 19 6959 9 11243 5 67.2 0
——Sequential Create—— ——–Random Create——–
-Create– –Read— -Delete– -Create– –Read— -Delete–
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 958 90 +++++ +++ +++++ +++ 1095 96 +++++ +++ 5052 94



Then I tested the SCSI disk (/) the command line was:


bonnie -d /home/tmp -s 1g -n master -r 512MB -u nobody:nogroup -f


And the results:




Version 1.02c ——Sequential Output—— –Sequential Input- –Random-
-Per Chr- –Block– -Rewrite- -Per Chr- –Block– –Seeks–
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
master 1G 19167 20 7908 6 16688 6 298.5 1
——Sequential Create—— ——–Random Create——–
-Create– –Read— -Delete– -Create– –Read— -Delete–
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 15163 99 +++++ +++ 12388 99 14929 100 +++++ +++ 7521 67



Basically from what I understand of bonnie, both disks (well, the SCSI vs the RAID) were pretty close, with each edging out the other on various tests.


Just for fun I also ran this test with the same parameters (except for the ram size) on this system, which has RAID0 SCSI, but old SCSI 🙂




Version 1.02c ——Sequential Output—— –Sequential Input- –Random-
-Per Chr- –Block– -Rewrite- -Per Chr- –Block– –Seeks–
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
naked.arcterex.n 1G 9458 45 4492 9 10433 9 89.5 1
——Sequential Create—— ——–Random Create——–
-Create– –Read— -Delete– -Create– –Read— -Delete–
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 296 96 +++++ +++ 19496 99 314 98 +++++ +++ 1199 80

3 Comments on “Tech Stuff”

  1. Maybe a short perl script that creates/reads/writes or something?
    I have an X15 in my workstation, and have been quite happy with it, even without a proper LVD controller.
    Another important point is the filesystem being used.

  2. Interesting.
    Hit StorageReview to compare any drive. Go to Database->Legacy tests, then compare the X15 with the Maxtor. This will give a good indicator the performance, although you can’t compare a single drive against a RAID of other drives.
    A buddy had his 30GB Maxtor pack it in (tons of corrupted sectors), so he is getting a 36GB X15.3… will be interesting to hear from him how much faster it will be then his RAID array of Atlas 10,000 rpm drive. Apparently a single X15.3 is 4-5 times faster then a two drive RAID array of SCSI 10,000 rpm drives.
    CPU usage and throughput speed is important, but don’t forget to check the seek times… on a file/web/mail server, seek times (how long it takes to position the head over a piece of data) are probably much more important then throughput. I’m not sure if Bonnie does this, but compare the results of running a multi-threaded/multi-process benchmark tool… ie, creating/modifying/deleting many files, in parallel. Typically scsi holds up much better when there are lots of concurrent reads/writes (such as 50 people retrieving web pages and doing database calls) then IDE does. So, a simple benchmark that imitates a single user(process) doing stuff isn’t always the most fair. Last night I was working on a system with dual PIII 850’s and a scsi raid array, and it definetly seemed very responsive.
    Damn. That was a lot of typing and probably makes no sense. If it sounds like Wim babbling on and on again, stating the obvious, sorry… Must sleep now.

  3. This is the thing I’d like to know…. What is the best way to test this. Reason I might be able to get my hands on 1 or more X15 disks and wonder if the performance advantage is worth the extra $$$.;) Ever seen a good test to really test it out. Perhaps just run a script that runs 10 versions of bonnie should do it.;)