(Revision) RAID performance

From: QPXu <xu_at_fleep.ocs.ou.edu>
Date: Thu Aug 30 2001 - 14:06:24 EDT
  This message is in MIME format.  The first part should be readable text,
  while the remaining parts are likely unreadable without MIME-aware tools.
  Send mail to mime@docserver.cac.washington.edu for more info.

--260141763-1482410743-999194784=:7691
Content-Type: TEXT/PLAIN; charset=US-ASCII


Based on responses I got, I have revised my original
post to provide some additional information.


Hi,

I am suspicious of the disk performance of our Sun A5100 RAID
(14*36GB disk with fiber). I have looked at the various mail-lists,
many people claimed that they can get at least 50-60MB/s
(max 100MB/s). I cannot see this kind of performance with
our RAID unit. For a rough estimate, by copying a large file,
on Solaris it took much longer (~3-5MB/s) than a PC (~10MB/s).

In order to get a better idea, I used bonnie (http://www.textuality.com/bonnie/)
as the benchmark tool, two systems are used in the test

  1. A5100 attached to two E3500s (solaris 8) each
     with dual path/fibre, this configuration put the RAID
     under the control of two Sun E3500, the controllers
     are shared, each E3500 control part of the RAID (disks
     cannot be shared). The software used to control
     the RAID is VXVM 3.1, filesystem is ufs. The parameters
     for making the filesystem are based on default. Each
     E3500 has 4x400Mhz CPUs and 1GB RAM.

     The disks used in our A5100 is: SEAGATE ST136403FSUN36G  Revision: 034A
        Not sure what kind of scsi it is. but is probably not LvM-160.

     The RAID is divided into multiple volumes, I have tested 4-disk
     strip-pro (RAID 1+0), single disk on RAID and non-RAID
     disk (disks come from E3500). I haven't bother to try RAID5
     since I would not expect it to beat RAID 1+0.


  2. A home-made PC
     ASUS A7V motherboard with 1GHz Athlon CPU +1.5GB PC133
     OS: FreeBSD 4.3
     There is two disks:
     	30GB ATA100 by maxtor
	18GB SCSI ultra160 by Quantum: Atlas V

The results of bonnie is at the end of the message and also
as an attachment (in case of formatting screw-up). It is
obvious that the PC is doing better in almost everything.

I have also benchmark the RAID unit with vxbench. It looks
like only if many processes are writting/reading at the same
time, A5100 will do better than single disk.

I have suspected that our setup/configurations were not
right/optimal. The overall system works fine with no apparent
hardware problems. So there must be something I am missing
here: either my setup is not optimal or A5100 doesn't
live up to its claims. Does anyone have any experiences
RAID setup/performance tuning/benchmarking ? It would be great
is you can share the performance data of the RAIDs
under your administration. Thanks ahead,

-Qingping Xu
Univ. of Oklahoma

Output generated by "bonnie -s 1000"
=====================================

FreeBSD PC SCSI ultra160

     -------Sequential Output-------- ---Sequential Input-- --Random--
     -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
  MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
1000 22106 39.0 20108 12.0 10133  8.6 19278 51.8 16873  9.2 240.0  1.4

FreeBSD PC IDE ATA66

     -------Sequential Output-------- ---Sequential Input-- --Random--
     -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
  MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
1000 25866 44.6 25546 14.7 11860 11.4 23107 61.9 26811 14.4 168.2  1.0

Solaris 8 A5100 4-disk strip-pro

       ---Sequential Output (nosync)--- ---Sequential Input-- --Rnd Seek-
       -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --04k (03)-
    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU   /sec %CPU
1*1000  5695 39.3  4323  9.7  3271 11.9 14122 82.9 26575 38.0  131.3 10.5

Solaris 8 A5100 single disk

       ---Sequential Output (nosync)--- ---Sequential Input-- --Rnd Seek-
       -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --04k (03)-
    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU   /sec %CPU
1*1000  6771 42.5  6751 10.7  5207 13.8 13197 76.6 20492 26.7   82.0  6.2


Solaris 8 system disk (nonRAID)

       ---Sequential Output (nosync)--- ---Sequential Input-- --Rnd Seek-
       -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --04k (03)-
    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU   /sec %CPU
1*1000  9533 57.8 15974 21.5  7517 15.6 12854 71.2 21452 23.5   40.8  3.2

--260141763-1482410743-999194784=:7691
Content-Type: TEXT/plain; name="b.txt"
Content-Transfer-Encoding: BASE64
Content-ID: <Pine.LNX.4.33.0108301306240.7691@fleep.ocs.ou.edu>
Content-Description: bonnie output summary
Content-Disposition: attachment; filename="b.txt"

T3V0cHV0IGdlbmVyYXRlZCBieSAiYm9ubmllIC1zIDEwMDAiDQo9PT09PT09
PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09DQoNCkZyZWVCU0QgUEMg
U0NTSSB1bHRyYTE2MA0KDQogICAgIC0tLS0tLS1TZXF1ZW50aWFsIE91dHB1
dC0tLS0tLS0tIC0tLVNlcXVlbnRpYWwgSW5wdXQtLSAtLVJhbmRvbS0tDQog
ICAgIC1QZXIgQ2hhci0gLS1CbG9jay0tLSAtUmV3cml0ZS0tIC1QZXIgQ2hh
ci0gLS1CbG9jay0tLSAtLVNlZWtzLS0tDQogIE1CIEsvc2VjICVDUFUgSy9z
ZWMgJUNQVSBLL3NlYyAlQ1BVIEsvc2VjICVDUFUgSy9zZWMgJUNQVSAgL3Nl
YyAlQ1BVDQoxMDAwIDIyMTA2IDM5LjAgMjAxMDggMTIuMCAxMDEzMyAgOC42
IDE5Mjc4IDUxLjggMTY4NzMgIDkuMiAyNDAuMCAgMS40DQoNCkZyZWVCU0Qg
UEMgSURFIEFUQTY2DQoNCiAgICAgLS0tLS0tLVNlcXVlbnRpYWwgT3V0cHV0
LS0tLS0tLS0gLS0tU2VxdWVudGlhbCBJbnB1dC0tIC0tUmFuZG9tLS0NCiAg
ICAgLVBlciBDaGFyLSAtLUJsb2NrLS0tIC1SZXdyaXRlLS0gLVBlciBDaGFy
LSAtLUJsb2NrLS0tIC0tU2Vla3MtLS0NCiAgTUIgSy9zZWMgJUNQVSBLL3Nl
YyAlQ1BVIEsvc2VjICVDUFUgSy9zZWMgJUNQVSBLL3NlYyAlQ1BVICAvc2Vj
ICVDUFUNCjEwMDAgMjU4NjYgNDQuNiAyNTU0NiAxNC43IDExODYwIDExLjQg
MjMxMDcgNjEuOSAyNjgxMSAxNC40IDE2OC4yICAxLjANCg0KU29sYXJpcyA4
IEE1MTAwIDQtZGlzayBzdHJpcC1wcm8NCg0KICAgICAgIC0tLVNlcXVlbnRp
YWwgT3V0cHV0IChub3N5bmMpLS0tIC0tLVNlcXVlbnRpYWwgSW5wdXQtLSAt
LVJuZCBTZWVrLQ0KICAgICAgIC1QZXIgQ2hhci0gLS1CbG9jay0tLSAtUmV3
cml0ZS0tIC1QZXIgQ2hhci0gLS1CbG9jay0tLSAtLTA0ayAoMDMpLQ0KICAg
IE1CIEsvc2VjICVDUFUgSy9zZWMgJUNQVSBLL3NlYyAlQ1BVIEsvc2VjICVD
UFUgSy9zZWMgJUNQVSAgIC9zZWMgJUNQVQ0KMSoxMDAwICA1Njk1IDM5LjMg
IDQzMjMgIDkuNyAgMzI3MSAxMS45IDE0MTIyIDgyLjkgMjY1NzUgMzguMCAg
MTMxLjMgMTAuNQ0KDQpTb2xhcmlzIDggQTUxMDAgc2luZ2xlIGRpc2sNCg0K
ICAgICAgIC0tLVNlcXVlbnRpYWwgT3V0cHV0IChub3N5bmMpLS0tIC0tLVNl
cXVlbnRpYWwgSW5wdXQtLSAtLVJuZCBTZWVrLQ0KICAgICAgIC1QZXIgQ2hh
ci0gLS1CbG9jay0tLSAtUmV3cml0ZS0tIC1QZXIgQ2hhci0gLS1CbG9jay0t
LSAtLTA0ayAoMDMpLQ0KICAgIE1CIEsvc2VjICVDUFUgSy9zZWMgJUNQVSBL
L3NlYyAlQ1BVIEsvc2VjICVDUFUgSy9zZWMgJUNQVSAgIC9zZWMgJUNQVQ0K
MSoxMDAwICA2NzcxIDQyLjUgIDY3NTEgMTAuNyAgNTIwNyAxMy44IDEzMTk3
IDc2LjYgMjA0OTIgMjYuNyAgIDgyLjAgIDYuMg0KDQoNClNvbGFyaXMgOCBz
eXN0ZW0gZGlzayAobm9uUkFJRCkNCg0KICAgICAgIC0tLVNlcXVlbnRpYWwg
T3V0cHV0IChub3N5bmMpLS0tIC0tLVNlcXVlbnRpYWwgSW5wdXQtLSAtLVJu
ZCBTZWVrLQ0KICAgICAgIC1QZXIgQ2hhci0gLS1CbG9jay0tLSAtUmV3cml0
ZS0tIC1QZXIgQ2hhci0gLS1CbG9jay0tLSAtLTA0ayAoMDMpLQ0KICAgIE1C
IEsvc2VjICVDUFUgSy9zZWMgJUNQVSBLL3NlYyAlQ1BVIEsvc2VjICVDUFUg
Sy9zZWMgJUNQVSAgIC9zZWMgJUNQVQ0KMSoxMDAwICA5NTMzIDU3LjggMTU5
NzQgMjEuNSAgNzUxNyAxNS42IDEyODU0IDcxLjIgMjE0NTIgMjMuNSAgIDQw
LjggIDMuMg0K
--260141763-1482410743-999194784=:7691--
Received on Thu Aug 30 19:06:24 2001

This archive was generated by hypermail 2.1.8 : Wed Mar 23 2016 - 16:25:03 EDT