Veritas-bu

[Veritas-bu] iostat during duplication - strange

2002-04-04 13:58:35
Subject: [Veritas-bu] iostat during duplication - strange
From: Mark.Donaldson AT experianems DOT com (Donaldson, Mark)
Date: Thu, 4 Apr 2002 11:58:35 -0700
This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.

------_=_NextPart_001_01C1DC0A.B92F8F10
Content-Type: text/plain

Update:

For those following the saga - I've done two changes to increase duplication
speed:

I've matched buffer sizes, NET_BUFFER_SZ & SIZE_DATA_BUFFERS.  This had a
measurable effect on my client restores, too, moving a test DB restore from
over 17 hours to just about 10 hours.

See: http://seer.support.veritas.com/docs/183702.htm
and: http://seer.support.veritas.com/docs/243197.htm

I'm also now duplicating from non-multiplexed images.  Before I was
demultiplexing to my duplicate set from a multiplexed set.  (It reduces
tape-drive usage during the dup.)

I've built another class for a select set of clients that isn't using
schedule multiplexing.  It's also using a different tape pool so other
schedules that kick off during the backup period won't try to mux at the
drive level.  This dup set is running now so the final results for are not
in yet but you can see the difference in the iostat output.

  r/s  w/s   kr/s   kw/s wait actv wsvc_t asvc_t  %w  %b device
 18.2  0.0 4667.1    0.0  0.0  1.0    0.0   54.6   0 100 rmt/0
  0.0 18.2    0.0 4667.1  0.0  0.4    0.0   23.3   0  42 rmt/4
  r/s  w/s   kr/s   kw/s wait actv wsvc_t asvc_t  %w  %b device
 16.6  0.0 4262.4    0.0  0.0  1.0    0.0   59.9   0 100 rmt/0
  0.0 16.7    0.0 4266.6  0.0  0.4    0.0   24.6   0  41 rmt/4
  r/s  w/s   kr/s   kw/s wait actv wsvc_t asvc_t  %w  %b device
 21.0  0.0 5380.2    0.0  0.0  1.0    0.0   47.3   0  99 rmt/0
  0.0 21.0    0.0 5371.7  0.0  0.5    0.0   24.3   0  51 rmt/4
  r/s  w/s   kr/s   kw/s wait actv wsvc_t asvc_t  %w  %b device
 21.6  0.0 5521.0    0.0  0.0  1.0    0.0   46.1   0  99 rmt/0
  0.0 21.6    0.0 5529.5  0.0  0.5    0.0   24.2   0  52 rmt/4
  r/s  w/s   kr/s   kw/s wait actv wsvc_t asvc_t  %w  %b device
 17.7  0.0 4531.0    0.0  0.0  1.0    0.0   55.6   0  99 rmt/0
  0.0 17.7    0.0 4526.7  0.0  0.4    0.0   24.8   0  44 rmt/4

As you can see, the kr/s pretty-much matches kw/s, something that was
25%-50% (depending on multiplexing factor) & less before when reading from
multiplexed sources.  I think these changes are going to prove to be my "big
solution" until the multi-write features of v4.5 arrive.

-Mark

------_=_NextPart_001_01C1DC0A.B92F8F10
Content-Type: text/html
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =
charset=3DUS-ASCII">
<META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version =
5.5.2653.12">
<TITLE>RE: [Veritas-bu] iostat during duplication - strange</TITLE>
</HEAD>
<BODY>

<P><FONT SIZE=3D2>Update:</FONT>
</P>

<P><FONT SIZE=3D2>For those following the saga - I've done two changes =
to increase duplication speed:</FONT>
</P>

<P><FONT SIZE=3D2>I've matched buffer sizes, NET_BUFFER_SZ &amp; =
SIZE_DATA_BUFFERS.&nbsp; This had a measurable effect on my client =
restores, too, moving a test DB restore from over 17 hours to just =
about 10 hours.</FONT></P>

<P><FONT SIZE=3D2>See: <A =
HREF=3D"http://seer.support.veritas.com/docs/183702.htm"; =
TARGET=3D"_blank">http://seer.support.veritas.com/docs/183702.htm</A></F=
ONT>
<BR><FONT SIZE=3D2>and: <A =
HREF=3D"http://seer.support.veritas.com/docs/243197.htm"; =
TARGET=3D"_blank">http://seer.support.veritas.com/docs/243197.htm</A></F=
ONT>
</P>

<P><FONT SIZE=3D2>I'm also now duplicating from non-multiplexed =
images.&nbsp; Before I was demultiplexing to my duplicate set from a =
multiplexed set.&nbsp; (It reduces tape-drive usage during the =
dup.)</FONT></P>

<P><FONT SIZE=3D2>I've built another class for a select set of clients =
that isn't using schedule multiplexing.&nbsp; It's also using a =
different tape pool so other schedules that kick off during the backup =
period won't try to mux at the drive level.&nbsp; This dup set is =
running now so the final results for are not in yet but you can see the =
difference in the iostat output.</FONT></P>

<P><FONT SIZE=3D2>&nbsp; r/s&nbsp; w/s&nbsp;&nbsp; kr/s&nbsp;&nbsp; =
kw/s wait actv wsvc_t asvc_t&nbsp; %w&nbsp; %b device</FONT>
<BR><FONT SIZE=3D2>&nbsp;18.2&nbsp; 0.0 4667.1&nbsp;&nbsp;&nbsp; =
0.0&nbsp; 0.0&nbsp; 1.0&nbsp;&nbsp;&nbsp; 0.0&nbsp;&nbsp; =
54.6&nbsp;&nbsp; 0 100 rmt/0</FONT>
<BR><FONT SIZE=3D2>&nbsp; 0.0 18.2&nbsp;&nbsp;&nbsp; 0.0 4667.1&nbsp; =
0.0&nbsp; 0.4&nbsp;&nbsp;&nbsp; 0.0&nbsp;&nbsp; 23.3&nbsp;&nbsp; =
0&nbsp; 42 rmt/4</FONT>
<BR><FONT SIZE=3D2>&nbsp; r/s&nbsp; w/s&nbsp;&nbsp; kr/s&nbsp;&nbsp; =
kw/s wait actv wsvc_t asvc_t&nbsp; %w&nbsp; %b device</FONT>
<BR><FONT SIZE=3D2>&nbsp;16.6&nbsp; 0.0 4262.4&nbsp;&nbsp;&nbsp; =
0.0&nbsp; 0.0&nbsp; 1.0&nbsp;&nbsp;&nbsp; 0.0&nbsp;&nbsp; =
59.9&nbsp;&nbsp; 0 100 rmt/0</FONT>
<BR><FONT SIZE=3D2>&nbsp; 0.0 16.7&nbsp;&nbsp;&nbsp; 0.0 4266.6&nbsp; =
0.0&nbsp; 0.4&nbsp;&nbsp;&nbsp; 0.0&nbsp;&nbsp; 24.6&nbsp;&nbsp; =
0&nbsp; 41 rmt/4</FONT>
<BR><FONT SIZE=3D2>&nbsp; r/s&nbsp; w/s&nbsp;&nbsp; kr/s&nbsp;&nbsp; =
kw/s wait actv wsvc_t asvc_t&nbsp; %w&nbsp; %b device</FONT>
<BR><FONT SIZE=3D2>&nbsp;21.0&nbsp; 0.0 5380.2&nbsp;&nbsp;&nbsp; =
0.0&nbsp; 0.0&nbsp; 1.0&nbsp;&nbsp;&nbsp; 0.0&nbsp;&nbsp; =
47.3&nbsp;&nbsp; 0&nbsp; 99 rmt/0</FONT>
<BR><FONT SIZE=3D2>&nbsp; 0.0 21.0&nbsp;&nbsp;&nbsp; 0.0 5371.7&nbsp; =
0.0&nbsp; 0.5&nbsp;&nbsp;&nbsp; 0.0&nbsp;&nbsp; 24.3&nbsp;&nbsp; =
0&nbsp; 51 rmt/4</FONT>
<BR><FONT SIZE=3D2>&nbsp; r/s&nbsp; w/s&nbsp;&nbsp; kr/s&nbsp;&nbsp; =
kw/s wait actv wsvc_t asvc_t&nbsp; %w&nbsp; %b device</FONT>
<BR><FONT SIZE=3D2>&nbsp;21.6&nbsp; 0.0 5521.0&nbsp;&nbsp;&nbsp; =
0.0&nbsp; 0.0&nbsp; 1.0&nbsp;&nbsp;&nbsp; 0.0&nbsp;&nbsp; =
46.1&nbsp;&nbsp; 0&nbsp; 99 rmt/0</FONT>
<BR><FONT SIZE=3D2>&nbsp; 0.0 21.6&nbsp;&nbsp;&nbsp; 0.0 5529.5&nbsp; =
0.0&nbsp; 0.5&nbsp;&nbsp;&nbsp; 0.0&nbsp;&nbsp; 24.2&nbsp;&nbsp; =
0&nbsp; 52 rmt/4</FONT>
<BR><FONT SIZE=3D2>&nbsp; r/s&nbsp; w/s&nbsp;&nbsp; kr/s&nbsp;&nbsp; =
kw/s wait actv wsvc_t asvc_t&nbsp; %w&nbsp; %b device</FONT>
<BR><FONT SIZE=3D2>&nbsp;17.7&nbsp; 0.0 4531.0&nbsp;&nbsp;&nbsp; =
0.0&nbsp; 0.0&nbsp; 1.0&nbsp;&nbsp;&nbsp; 0.0&nbsp;&nbsp; =
55.6&nbsp;&nbsp; 0&nbsp; 99 rmt/0</FONT>
<BR><FONT SIZE=3D2>&nbsp; 0.0 17.7&nbsp;&nbsp;&nbsp; 0.0 4526.7&nbsp; =
0.0&nbsp; 0.4&nbsp;&nbsp;&nbsp; 0.0&nbsp;&nbsp; 24.8&nbsp;&nbsp; =
0&nbsp; 44 rmt/4</FONT>
</P>

<P><FONT SIZE=3D2>As you can see, the kr/s pretty-much matches kw/s, =
something that was 25%-50% (depending on multiplexing factor) &amp; =
less before when reading from multiplexed sources.&nbsp; I think these =
changes are going to prove to be my &quot;big solution&quot; until the =
multi-write features of v4.5 arrive.</FONT></P>

<P><FONT SIZE=3D2>-Mark</FONT>
</P>

</BODY>
</HTML>
------_=_NextPart_001_01C1DC0A.B92F8F10--