144 lines
8.6 KiB
HTML
144 lines
8.6 KiB
HTML
|
<?xml version="1.0" encoding="utf-8"?>
|
||
|
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
|
||
|
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
|
||
|
<html xmlns="http://www.w3.org/1999/xhtml" lang="en-US" xml:lang="en-us">
|
||
|
<head>
|
||
|
<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
|
||
|
<meta name="dc.language" scheme="rfc1766" content="en-us" />
|
||
|
<!-- All rights reserved. Licensed Materials Property of IBM -->
|
||
|
<!-- US Government Users Restricted Rights -->
|
||
|
<!-- Use, duplication or disclosure restricted by -->
|
||
|
<!-- GSA ADP Schedule Contract with IBM Corp. -->
|
||
|
<meta name="dc.date" scheme="iso8601" content="2005-10-03" />
|
||
|
<meta name="copyright" content="(C) Copyright IBM Corporation 1998, 2006" />
|
||
|
<meta name="security" content="public" />
|
||
|
<meta name="Robots" content="index,follow"/>
|
||
|
<meta http-equiv="PICS-Label" content='(PICS-1.1 "http://www.icra.org/ratingsv02.html" l gen true r (cz 1 lz 1 nz 1 oz 1 vz 1) "http://www.rsac.org/ratingsv01.html" l gen true r (n 0 s 0 v 0 l 0) "http://www.classify.org/safesurf/" l gen true r (SS~~000 1))' />
|
||
|
<title>Performance considerations for APPN and HPR</title>
|
||
|
<link rel="stylesheet" type="text/css" href="ibmidwb.css" />
|
||
|
<link rel="stylesheet" type="text/css" href="ic.css" />
|
||
|
</head>
|
||
|
<body>
|
||
|
<a id="Top_Of_Page" name="Top_Of_Page"></a><!-- Java sync-link -->
|
||
|
<script language = "Javascript" src = "../rzahg/synch.js" type="text/javascript"></script>
|
||
|
|
||
|
|
||
|
<a name="rzahjper-comappnfacts"></a>
|
||
|
<h3 id="rzahjper-comappnfacts">Performance considerations for APPN and HPR</h3>
|
||
|
<p>The following can affect the performance of the APPN and HPR protocols:</p>
|
||
|
<ul>
|
||
|
<li>Transmission priority
|
||
|
<p>When you create a class-of-service description,
|
||
|
you can define one of three transmission priorities for each class of service.
|
||
|
You can specify, using the transmission priority (TMSPTY) parameter, that
|
||
|
the transmission priority for any class of service is high, medium, or low.</p>
|
||
|
<p>The transmission priority you specify is carried in the session activation
|
||
|
request at session establishment. The transmission priority allows each logical
|
||
|
unit on the session and each routing entry along the session path to store
|
||
|
the same transmission priority. By assigning an appropriate mode (which includes
|
||
|
a class of service) at session activation time, you can ensure a better response
|
||
|
time for the applications that require it. Generally, interactive traffic
|
||
|
should have a high priority and batch traffic a low priority.</p></li>
|
||
|
<li>Route addition resistance
|
||
|
<p>Route addition resistance (RAR) is a relative
|
||
|
value that indicates how desirable one network node is, as compared to other
|
||
|
network nodes, for having intermediate sessions routed through it.</p>
|
||
|
<p>Changing this value and working with the different class-of-service descriptions
|
||
|
can control route sessions.</p>
|
||
|
<p>The RAR value is defined in the network
|
||
|
attributes for the local iSeries system.</p></li>
|
||
|
<li>Pacing values: See <a href="rzahjperpacingfacts.htm#rzahjper-pacingfacts">Pacing (INPACING,
|
||
|
OUTPACING, MAXINPACING) parameters</a> for pacing considerations.</li>
|
||
|
<li>Session activation considerations
|
||
|
<p>When a session is requested to a remote
|
||
|
location that matches a network node control-point name, a directory search
|
||
|
is not performed by the node that calculates the route. This is true if the
|
||
|
session request is being started by a user on the network node, or on an end
|
||
|
node that the network node is providing services for. Session start requests
|
||
|
for remote locations in end nodes and remote locations in network nodes that
|
||
|
do not match the control-point name of the network nodes take longer. These
|
||
|
session start requests take longer because the directory search needs to be
|
||
|
sent and the replies need to be received.</p></li>
|
||
|
<li>Maximum intermediate sessions
|
||
|
<p>The Change Network Attributes (CHGNETA)
|
||
|
command specifies the maximum number of intermediate sessions that are allowed
|
||
|
on a network node. When the number of intermediate sessions reaches 90% of
|
||
|
the maximum value, the node is marked as congested. A node that is congested
|
||
|
may or may not be used for intermediate sessions that depend on the class-of-service
|
||
|
definition. The node is not congested when the number of intermediate sessions
|
||
|
drops below 80% of the configured value. Also, if the maximum number of intermediate
|
||
|
sessions is reached (100%), then intermediate sessions will not be allowed
|
||
|
through this network node until the value drops. You can limit the effect
|
||
|
of intermediate sessions on local processing by setting an appropriate value.</p></li>
|
||
|
<li>Segmentation and reassembly
|
||
|
<p>On iSeries™, some IOPs that support the local area
|
||
|
network protocols, such as token ring and Ethernet, have the ability to perform
|
||
|
segmentation and reassembly of SNA Request Units. Performing this function
|
||
|
in the IOPs offloads this work from the server CPU. The server CPU is free
|
||
|
to perform other tasks.</p>
|
||
|
<a name="wq89"></a>
|
||
|
<div class="notetitle" id="wq89">Note:</div><img src="delta.gif" alt="Start of change" />
|
||
|
<div class="notebody">Communications input/output
|
||
|
adapters (IOAs), such as Gigabit Ethernet, do not require an input/output
|
||
|
processor (IOP) and therefore perform segmentation in the server CPU. Gigabit
|
||
|
Ethernet adapters do not automatically support SNA. Enterprise Extender (or AnyNet®) is required to allow SNA data to flow over a Gigabit adapter. IBM® recommends
|
||
|
that Enterprise Extender be used in place of AnyNet.</div><img src="deltaend.gif" alt="End of change" />
|
||
|
<p><img src="delta.gif" alt="Start of change" />With
|
||
|
APPN, any network congestion control is handled on a hop-by-hop basis by using <a href="rzahjperpacingfacts.htm#rzahjper-pacingfacts">pacing </a> values. It is possible to over-drive
|
||
|
connections in an APPN environment. A particular system might receive more
|
||
|
data over a communications link than it can handle based on buffer space.
|
||
|
The system requires the node that sends the data to retransmit all of the
|
||
|
frames that were sent following the last successfully acknowledged frame.
|
||
|
This retransmission occurs at the data link control (DLC) layer.</p>
|
||
|
<a name="wq90"></a>
|
||
|
<div class="notetitle" id="wq90">Note:</div>
|
||
|
<div class="notebody">HPR has little IOP assistance. Much of the segmentation and reassembly is
|
||
|
done in the server CPU.</div><img src="deltaend.gif" alt="End of change" /></li>
|
||
|
<li>Error recovery
|
||
|
<p>APPN requires link-level error recovery to cause retransmission
|
||
|
of lost frames. This link-level error recovery can only survive short and
|
||
|
temporary outages (several seconds). If a link outage or node outage occurs,
|
||
|
that is of longer duration, APPN has no recovery mechanisms for keeping the
|
||
|
affected sessions active. The applications must handle any session recovery.</p>
|
||
|
<p>The following matrix shows how HPR traffic is supported between two
|
||
|
systems that are based on their HPR link-level error recovery settings. The
|
||
|
HPR link-level error settings are exchanged between the systems: </p><img src="delta.gif" alt="Start of change" />
|
||
|
<a name="wq91"></a>
|
||
|
<table id="wq91" width="90%" summary="" border="1" frame="border" rules="all">
|
||
|
<tbody valign="top">
|
||
|
<tr>
|
||
|
<td width="29%" rowspan="2" align="left" valign="top"><span class="bold">System 1</span></td>
|
||
|
<td colspan="3" align="left" valign="top"><span class="bold">System 2</span></td>
|
||
|
</tr>
|
||
|
<tr>
|
||
|
<td width="24%" align="left" valign="top">No link-level ERP allowed</td>
|
||
|
<td width="23%" align="left" valign="top">Link-level ERP required</td>
|
||
|
<td width="22%" align="left" valign="top">Prefer no link-level ERP but might run using
|
||
|
link-level ERP</td>
|
||
|
</tr>
|
||
|
<tr>
|
||
|
<td align="left" valign="top">No link-level ERP allowed</td>
|
||
|
<td align="left" valign="top">HPR supported (no ERP)</td>
|
||
|
<td align="left" valign="top">HPR not used</td>
|
||
|
<td align="left" valign="top">HPR supported (no ERP)</td>
|
||
|
</tr>
|
||
|
<tr>
|
||
|
<td align="left" valign="top">Link-level ERP required</td>
|
||
|
<td align="left" valign="top">HPR not used</td>
|
||
|
<td align="left" valign="top">HPR supported (uses ERP)</td>
|
||
|
<td align="left" valign="top">HPR supported (uses ERP)</td>
|
||
|
</tr>
|
||
|
<tr>
|
||
|
<td align="left" valign="top">Prefer no link-level ERP but might run using
|
||
|
link-level ERP</td>
|
||
|
<td align="left" valign="top">HPR supported (no ERP)</td>
|
||
|
<td align="left" valign="top">HPR supported (uses ERP)</td>
|
||
|
<td align="left" valign="top">HPR supported (no ERP)</td>
|
||
|
</tr>
|
||
|
</tbody>
|
||
|
</table><img src="deltaend.gif" alt="End of change" /></li></ul>
|
||
|
<p>For information about high-performance routing, see <a href="rzahjpercomhprfacts.htm#rzahjper-comhprfacts">Optimize communications using high-performance routing</a>.</p>
|
||
|
<a id="Bot_Of_Page" name="Bot_Of_Page"></a>
|
||
|
</body>
|
||
|
</html>
|