2.4.x/docs/manual/platform/perf-hp.xml

<?xml version="1.0" encoding="UTF-8" ?>
<!DOCTYPE manualpage SYSTEM "../style/manualpage.dtd">
<?xml-stylesheet type="text/xsl" href="../style/manual.en.xsl"?>
<!-- $LastChangedRevision$ -->

<!--
 Licensed to the Apache Software Foundation (ASF) under one or more
 contributor license agreements.  See the NOTICE file distributed with
 this work for additional information regarding copyright ownership.
 The ASF licenses this file to You under the Apache License, Version 2.0
 (the "License"); you may not use this file except in compliance with
 the License.  You may obtain a copy of the License at

     http://www.apache.org/licenses/LICENSE-2.0

 Unless required by applicable law or agreed to in writing, software
 distributed under the License is distributed on an "AS IS" BASIS,
 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License.
-->

<manualpage metafile="perf-hp.xml.meta">
  <parentdocument href="./">Platform Specific Notes</parentdocument>

  <title>Running a High-Performance Web Server on HPUX</title>

  <summary>

<pre>
Date: Wed, 05 Nov 1997 16:59:34 -0800
From: Rick Jones &lt;<a
href="mailto:raj@cup.hp.com">raj@cup.hp.com</a>&gt;
Reply-To: <a href="mailto:raj@cup.hp.com"
>raj@cup.hp.com</a>
Organization: Network Performance
Subject: HP-UX tuning tips
</pre>

    <p>Here are some tuning tips for HP-UX to add to the tuning page.</p>

    <p>For HP-UX 9.X: Upgrade to 10.20<br />
    For HP-UX 10.[00|01|10]: Upgrade to 10.20</p>

    <p>For HP-UX 10.20:</p>

    <p>Install the latest cumulative ARPA Transport Patch. This
    will allow you to configure the size of the TCP connection
    lookup hash table. The default is 256 buckets and must be set
    to a power of two. This is accomplished with adb against the
    *disc* image of the kernel. The variable name is <code>tcp_hash_size</code>.
    Notice that it's critically important that you use "<code>W</code>"
    to write a 32 bit quantity, not "<code>w</code>" to write a 16 bit
    value when patching the disc image because the <code>tcp_hash_size</code>
    variable is a 32 bit quantity.</p>

    <p>How to pick the value? Examine the output of <a
    href="ftp://ftp.cup.hp.com/dist/networking/tools/connhist"
    >ftp://ftp.cup.hp.com/dist/networking/tools/connhist</a>
    and see how many total TCP connections exist on the system. You
    probably want that number divided by the hash table size to be
    reasonably small, say less than 10. Folks can look at HP's
    SPECweb96 disclosures for some common settings. These can be
    found at <a
    href="http://www.specbench.org/">http://www.specbench.org/</a>.
    If an HP-UX system was performing at 1000 SPECweb96 connections
    per second, the <code>TIME_WAIT</code> time of 60 seconds would mean
    60,000 TCP "connections" being tracked.</p>

    <p>Folks can check their listen queue depths with <a
    href="ftp://ftp.cup.hp.com/dist/networking/misc/listenq"
    >ftp://ftp.cup.hp.com/dist/networking/misc/listenq</a>.</p>

    <p>If folks are running Apache on a PA-8000 based system, they
    should consider "chatr'ing" the Apache executable to have a
    large page size. This would be "<code>chatr +pi L &lt;BINARY&gt;</code>".
    The GID of the running executable must have <code>MLOCK</code> privileges.
    <code>Setprivgrp(1m)</code> should be consulted for assigning
    <code>MLOCK</code>. The change can be validated by running Glance
    and examining the memory regions of the server(s) to make sure that
    they show a non-trivial fraction of the text segment being locked.</p>

    <p>If folks are running Apache on MP systems, they might
    consider writing a small program that uses <code>mpctl()</code>
    to bind processes to processors. A simple <code>pid % numcpu</code>
    algorithm is probably sufficient. This might even go into the
    source code.</p>

    <p>If folks are concerned about the number of <code>FIN_WAIT_2</code>
    connections, they can use nettune to shrink the value of
    <code>tcp_keepstart</code>. However, they should be careful there -
    certainly do not make it less than oh two to four minutes. If
    <code>tcp_hash_size</code> has been set well, it is probably OK to
    let the <code>FIN_WAIT_2</code>'s take longer to timeout (perhaps
    even the default two hours) - they will not on average have a big
    impact on performance.</p>

    <p>There are other things that could go into the code base, but
    that might be left for another email. Feel free to drop me a
    message if you or others are interested.</p>

    <p>sincerely,</p>

    <p>rick jones</p>

    <p><a href="http://www.netperf.org/netperf/"
    >http://www.netperf.org/netperf/</a></p>

  </summary>

</manualpage>