X-MimeOLE: Produced By Microsoft Exchange V6.5
Received: by onstor-exch02.onstor.net 
	id <01C89B4F.2288A860@onstor-exch02.onstor.net>; Thu, 10 Apr 2008 14:09:16 -0700
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Content-class: urn:content-classes:message
Subject: RE: IOzone shows better performance with Sun NFS server than 2280. Why?
Date: Thu, 10 Apr 2008 14:09:13 -0700
Message-ID: <BB375AF679D4A34E9CA8DFA650E2B04E0959686D@onstor-exch02.onstor.net>
In-Reply-To: <BB375AF679D4A34E9CA8DFA650E2B04E09596591@onstor-exch02.onstor.net>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: IOzone shows better performance with Sun NFS server than 2280. Why?
Thread-Index: AcibEDWBta4638SURxm5UboSIe7xYgAPWqEA
References: <BB375AF679D4A34E9CA8DFA650E2B04E09596591@onstor-exch02.onstor.net>
From: "Yasuhiro Murata" <yasuhiro.murata@onstor.com>
To: "dl-cstech" <dl-cstech@onstor.com>

Hi All.

Does anybody have some thought on this?
Why SUN-SUN show better result than SUN-Bobcat on the larger files or
larger record size?

Compare the both results for the large files that have more than 32768KB
size.

Regards,
Yasuhiro Murata

-----Original Message-----
From: Yasuhiro Murata=20
Sent: Thursday, April 10, 2008 10:39 PM
To: dl-cstech
Subject: IOzone shows better performance with Sun NFS server than 2280.
Why?

Hello,

I'm in the performance test now for Hitachi's ULSI deal.
I need help to make the deal move forward and win.

Using IOzone, Sun(Sol10) based NFS server shows better result than
bobcat2280.
NFS clinet is a sun Solaris8.
Write performance for large file is forcused now.

I need to show soon why both result are so different.
Could anybody shed the light on this?

[Additinal information]
  - Traffic is only from one sun(Sol8) client.
    And only one 1Gbit(PCI bus) ethernet interface is used at the
client.
  - Backend storage is not a bottleneck.
  - CPU usage of bobcat is low.=20
    I don't think internal process of the bobcat is stuck.=20
  - CPU usage of the client(Sol8) keep 100% almost all time when NFS
server is bobcat.
    On the otherhand, CPU is not so high (avarage 40%) when server is
Sun(Sol10).
  - When HP(HP-UX11i) is the client we saw much better resut with the
same test.=20

  *** I have a reference data for "stats show ldavg",  "evm show stats",
and KPI if needed. ***

[Result summary]
   Followings are the results for both. (Whole results are attached.)
   What I wonder from the results is that SUN-SUN shows over 125MB/Sec
throughput for very large file write.
   I think these numbers are over single GigaBit Ethernet capability.
   (I think this may be the key point to show them the reason)

   I also attach the output of the "nfsstat -c" at the client for both
results.
   There seems not to be big difference in the number of NFS calls
between both. =20

=3D=3D=3D=3D=3D SUN-SUN result =3D=3D=3D=3D=3D
"Writer report"
        "4"  "8"  "16"  "32"  "64"  "128"  "256"  "512"  "1024"  "2048"
"4096"  "8192"  "16384"
"64"   45483  57507  56089  55269  51161=20
"128"   54214  58287  54077  54985  48468  68515=20
"256"   56501  61420  63936  59812  52352  62242  67155=20
"512"   51953  61236  56480  58972  56462  60051  64134  65396=20
"1024"   54011  57190  59149  59313  60566  60763  62565  66497  64184=20
"2048"   55495  56134  58932  57693  48790  60769  63476  59552  61936
64972=20
"4096"   53728  58082  57954  58437  56574  59974  61521  63474  61901
62325  63080=20
"8192"   53782  57199  58708  58756  57101  61668  62019  61731  64682
83267  95962  91096=20
"16384"   54730  57022  55949  58165  58585  62567  61739  60139  65801
62216  63243  65748  149098=20
"32768"   0  0  0  0  68818  62028  79324  74633  78486  75031  81572
78433  71314=20
"65536"   0  0  0  0  95465  106387  113351  116766  110811  116677
115744  106830  109709=20
"131072"   0  0  0  0  113874  115111  131291  123556  162515  146297
145752  147608  140084=20
"262144"   0  0  0  0  137521  148231  152321  147639  145697  148468
144791  144181  141262=20
"524288"   0  0  0  0  152937  157625  145202  137820  145130  147049
143439  145229  157250=20
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D

=3D=3D=3D=3D=3D SUN-2280 result =3D=3D=3D=3D=3D
"Writer report"
        "4"  "8"  "16"  "32"  "64"  "128"  "256"  "512"  "1024"  "2048"
"4096"  "8192"  "16384"
"64"   45452  57817  61303  59090  61303=20
"128"   55554  54561  58420  55942  56615  61274=20
"256"   54560  58987  59992  63132  57114  57685  63381=20
"512"   56456  61524  59280  59418  58056  61062  63896  61272=20
"1024"   51582  59044  58437  59313  58504  65020  63677  65890  66814=20
"2048"   53874  59436  62421  59209  59250  62559  61219  66556  65017
68277=20
"4096"   55341  59243  60236  60074  62439  62592  62393  61633  66223
67038  69050=20
"8192"   59961  61707  60987  61877  60513  63234  61677  63763  65095
67172  65724  63475=20
"16384"   56222  57461  58976  57886  57037  58362  59735  57241  63956
63643  64146  65706  65503=20
"32768"   0  0  0  0  56222  60973  59730  61951  62925  67390  64756
66575  64325=20
"65536"   0  0  0  0  56829  60373  60452  61835  63860  65021  86818
63599  65621=20
"131072"   0  0  0  0  56493  61143  60574  60968  64441  70816  64405
159111  65664=20
"262144"   0  0  0  0  58089  80410  75901  105358  188374  162977
166572  171275  174136=20
"524288"   0  0  0  0  94962  60200  61219  62017  63196  70994  74853
64975  63896=20
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D

