Re: Sparc has slow disk access
Several things. One, you left out a ton of information, as this
could be the result of many, many things. Where to start?
What kind of disk/controller for each machine/disk?
What is the results you get from running bonnie?
Can you give the output of hdparm for each disk?
Also notice the the x86 box you are talking about is a considerably
faster processor/memory combination, and no, I'm not talking about
bogomips.
Try the solaris test again, Binkey, and this time use a filesystem
other than /tmp for the output. Try setting up a ufs-log file
system and see if the performance doesn't exceed that of linux.
That's your homework for today. Tomorrow we'll get to advanced disk
performance considerations. ~:^)
a
"Tille, Andreas" wrote:
>
> Hello,
>
> I´ve got a Sparc E250 server and I´m runing Debian GNU/Linux on it.
> I now started to make some performance comparisons against a default PC.
> Here are the configurations I used:
>
> E250:
> ~> cat /proc/cpuinfo
> cpu : TI UltraSparc II (BlackBird)
> fpu : UltraSparc II integrated FPU
> promlib : Version 3 Revision 16
> prom : 3.16.1
> type : sun4u
> ncpus probed : 2
> ncpus active : 2
> Cpu0Bogo : 799.53
> Cpu1Bogo : 799.53
> MMU Type : Spitfire
> State:
> CPU0: online
> CPU1: online
>
> 2 GB RAM
> 4x36 GB SCSI
>
> ~> uname -a
> Linux bse 2.2.18pre21 #1 SMP Tue Dec 19 22:53:19 CET 2000 sparc64 unknown
>
> PC:
> ~> cat /proc/cpuinfo
> processor : 0
> vendor_id : GenuineIntel
> cpu family : 6
> model : 8
> model name : Pentium III (Coppermine)
> stepping : 6
> cpu MHz : 795.576
> cache size : 256 KB
> fpu : yes
> fpu_exception : yes
> cpuid level : 2
> wp : yes
> flags : fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat pse36 mmx fxsr sse
> bogomips : 1589.24
>
> 256 MB RAM
> 20 GB IDE
>
> With this setup I noticed that the PC is about two times faster in
> inserting data into a postgresql database. I´ve got a hint to test
> disk speed and I did so with the following script:
>
> #!/bin/bash
> date
> ls -l /home/tillea/hd-test.file
> i=0
> while [ $i -lt 100 ] ; do
> echo -n "$i "
> let i="$i+1"
> dd if=/home/tillea/hd-test.file of=/tmp/hd-test.file 2> /dev/null
> done
> ls -l /tmp/hd-test.file
> date
> ### end of scrip
>
> Here are the results:
>
> PC:
>
> $ uname -a
> Linux wr-linux02 2.4.5 #1 Mit Mai 30 08:24:06 CEST 2001 i686 unknown
>
> $ hd-test
> Don Aug 23 12:21:56 CEST 2001
> -rw-r--r-- 1 tillea admin 25600000 23. Aug 11:50 /home/tillea/hd-test.file
> 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 -rw-r--r-- 1 tillea admin 25600000 23. Aug 12:22 /tmp/hd-test.file
> Don Aug 23 12:22:47 CEST 2001
>
> ----> 51s for the PC
>
> Sparc:
>
> $ uname -a
> SunOS bse 5.8 Generic_108528-08 sun4u sparc SUNW,Ultra-250
>
> $ hd-test
> Thu Aug 23 13:23:38 MEST 2001
> -rw-r--r-- 1 tillea admin 25600000 Aug 23 13:13 /home/tillea/hd-test.file
> 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 -rw-r--r-- 1 tillea admin 25600000 Aug 23 13:26 /tmp/hd-test.file
> Thu Aug 23 13:26:49 MEST 2001
>
> -----> 191s for the Sparc running Solaris 8
>
> $ uname -a
> Linux bse 2.2.18pre21 #1 SMP Tue Dec 19 22:53:19 CET 2000 sparc64 unknown
>
> Don Aug 23 13:55:37 CEST 2001
> -rw-r--r-- 1 tillea admin 25600000 23. Aug 11:50 /home/tillea/hd-test.file
> 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 -rw-r--r-- 1 tillea admin 25600000 23. Aug 13:57 /tmp/hd-test.file
> Don Aug 23 13:57:16 CEST 2001
>
> -----> 99s for tge Sparc running Debian GNU/Linux
>
> So the relation is about:
>
> PC/linux : Sparc/Solaris : Sparc/Linux
>
> 1 : 3.7 : 1.9
>
> That´s not good for the Sparc machine (and even worse for Solaris by the way
> but this is of less interest because the machine should run under Linux).
>
> Any hint if there some optimisations could be done?
Reply to: