<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
	xmlns:georss="http://www.georss.org/georss" xmlns:geo="http://www.w3.org/2003/01/geo/wgs84_pos#" xmlns:media="http://search.yahoo.com/mrss/"
		>
<channel>
	<title>Comments on: System Statistics</title>
	<atom:link href="http://jonathanlewis.wordpress.com/2010/12/17/system-statistics-2/feed/" rel="self" type="application/rss+xml" />
	<link>http://jonathanlewis.wordpress.com/2010/12/17/system-statistics-2/</link>
	<description>Just another Oracle weblog</description>
	<lastBuildDate>Thu, 23 May 2013 12:47:17 +0000</lastBuildDate>
	<sy:updatePeriod>hourly</sy:updatePeriod>
	<sy:updateFrequency>1</sy:updateFrequency>
	<generator>http://wordpress.com/</generator>
	<item>
		<title>By: Jonathan Lewis</title>
		<link>http://jonathanlewis.wordpress.com/2010/12/17/system-statistics-2/#comment-38829</link>
		<dc:creator><![CDATA[Jonathan Lewis]]></dc:creator>
		<pubDate>Wed, 22 Dec 2010 12:10:25 +0000</pubDate>
		<guid isPermaLink="false">http://jonathanlewis.wordpress.com/?p=4949#comment-38829</guid>
		<description><![CDATA[Kumar,

I&#039;m not entirely sure of the point you want to make, but I think by &quot;this argument&quot; you mean the comment on the metalink note. Two things then: (a) the fact that the metalink note says it is a good argument for EBS is not the same as saying it&#039;s a bad argument for everything else and (b) the blog I referenced has already made the point that I think that gathering system statistics is generally a bad idea.

Be careful that you distinguish between &quot;gathering&quot; system statistics and &quot;using&quot; system statistics. From 10g onwards you use system statistics by default, even if you haven&#039;t gathered them.

The dev/qa question is interesting - if you are trying to check the probable performance of the production code on a box that is significantly different from the test box then you have to worry about much more than just the system statistics: but it make sense to set them to the values used on the production box if you want to get the best chance of consistent execution plans.]]></description>
		<content:encoded><![CDATA[<p>Kumar,</p>
<p>I&#8217;m not entirely sure of the point you want to make, but I think by &#8220;this argument&#8221; you mean the comment on the metalink note. Two things then: (a) the fact that the metalink note says it is a good argument for EBS is not the same as saying it&#8217;s a bad argument for everything else and (b) the blog I referenced has already made the point that I think that gathering system statistics is generally a bad idea.</p>
<p>Be careful that you distinguish between &#8220;gathering&#8221; system statistics and &#8220;using&#8221; system statistics. From 10g onwards you use system statistics by default, even if you haven&#8217;t gathered them.</p>
<p>The dev/qa question is interesting &#8211; if you are trying to check the probable performance of the production code on a box that is significantly different from the test box then you have to worry about much more than just the system statistics: but it make sense to set them to the values used on the production box if you want to get the best chance of consistent execution plans.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Kumar</title>
		<link>http://jonathanlewis.wordpress.com/2010/12/17/system-statistics-2/#comment-38816</link>
		<dc:creator><![CDATA[Kumar]]></dc:creator>
		<pubDate>Tue, 21 Dec 2010 12:06:36 +0000</pubDate>
		<guid isPermaLink="false">http://jonathanlewis.wordpress.com/?p=4949#comment-38816</guid>
		<description><![CDATA[Hi Jonathan
Thanks for the reply.
But this argument is true for any application (not just oracle apps). In that case, is not the use of system statistics not justified in shops which have different environments (and the dev and qa boxes would not be on the same type of storage or speed as production boxes).

 - Kumar]]></description>
		<content:encoded><![CDATA[<p>Hi Jonathan<br />
Thanks for the reply.<br />
But this argument is true for any application (not just oracle apps). In that case, is not the use of system statistics not justified in shops which have different environments (and the dev and qa boxes would not be on the same type of storage or speed as production boxes).</p>
<p> &#8211; Kumar</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Jonathan Lewis</title>
		<link>http://jonathanlewis.wordpress.com/2010/12/17/system-statistics-2/#comment-38811</link>
		<dc:creator><![CDATA[Jonathan Lewis]]></dc:creator>
		<pubDate>Tue, 21 Dec 2010 07:59:47 +0000</pubDate>
		<guid isPermaLink="false">http://jonathanlewis.wordpress.com/?p=4949#comment-38811</guid>
		<description><![CDATA[See my reply to John Hurley.

If you follow my &quot;further reading&quot; links you&#039;ll note in the &quot;System Stats Strategy&quot; that  I&#039;ve made similar comments.

Possibly the SR makes the point it does because it is a benefit to the support team if they don&#039;t have to factor in different values for system stats whenever they&#039;re trying to work out what&#039;s going wrong with a client&#039;s query.]]></description>
		<content:encoded><![CDATA[<p>See my reply to John Hurley.</p>
<p>If you follow my &#8220;further reading&#8221; links you&#8217;ll note in the &#8220;System Stats Strategy&#8221; that  I&#8217;ve made similar comments.</p>
<p>Possibly the SR makes the point it does because it is a benefit to the support team if they don&#8217;t have to factor in different values for system stats whenever they&#8217;re trying to work out what&#8217;s going wrong with a client&#8217;s query.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Jonathan Lewis</title>
		<link>http://jonathanlewis.wordpress.com/2010/12/17/system-statistics-2/#comment-38810</link>
		<dc:creator><![CDATA[Jonathan Lewis]]></dc:creator>
		<pubDate>Tue, 21 Dec 2010 07:57:17 +0000</pubDate>
		<guid isPermaLink="false">http://jonathanlewis.wordpress.com/?p=4949#comment-38810</guid>
		<description><![CDATA[Joel.

Thanks for the link.]]></description>
		<content:encoded><![CDATA[<p>Joel.</p>
<p>Thanks for the link.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Jonathan Lewis</title>
		<link>http://jonathanlewis.wordpress.com/2010/12/17/system-statistics-2/#comment-38809</link>
		<dc:creator><![CDATA[Jonathan Lewis]]></dc:creator>
		<pubDate>Tue, 21 Dec 2010 07:56:32 +0000</pubDate>
		<guid isPermaLink="false">http://jonathanlewis.wordpress.com/?p=4949#comment-38809</guid>
		<description><![CDATA[John,

Quick summary (slightly evolved from the &quot;Further Posts&quot; due to the passage of time).
&lt;ul&gt;
Initial testing - from 10g onwards, don&#039;t set them
If necessary - use set_system_stats to create realistic seek and transfer figures
&lt;ul&gt;

Similarly, don&#039;t set db_file_multiblock_read_count, and don&#039;t set the two optimizer_ind% parameters (without extreme provocation.)]]></description>
		<content:encoded><![CDATA[<p>John,</p>
<p>Quick summary (slightly evolved from the &#8220;Further Posts&#8221; due to the passage of time).</p>
<ul>
Initial testing &#8211; from 10g onwards, don&#8217;t set them<br />
If necessary &#8211; use set_system_stats to create realistic seek and transfer figures
</ul>
<ul>
<p>Similarly, don&#8217;t set db_file_multiblock_read_count, and don&#8217;t set the two optimizer_ind% parameters (without extreme provocation.)</ul>
]]></content:encoded>
	</item>
	<item>
		<title>By: Jonathan Lewis</title>
		<link>http://jonathanlewis.wordpress.com/2010/12/17/system-statistics-2/#comment-38808</link>
		<dc:creator><![CDATA[Jonathan Lewis]]></dc:creator>
		<pubDate>Tue, 21 Dec 2010 07:51:54 +0000</pubDate>
		<guid isPermaLink="false">http://jonathanlewis.wordpress.com/?p=4949#comment-38808</guid>
		<description><![CDATA[Oleksander,

Thank you for supplying  the warning.

My ability to read Russian is limited to pressing buttons on Google Translate - and assuming that the occasional strange word is the wrong  literal translation of a word that can be used in many circumstances. Within the bounds that limitation, it seemed that most of the text made sense - but I hope readers following will be duly cautious.]]></description>
		<content:encoded><![CDATA[<p>Oleksander,</p>
<p>Thank you for supplying  the warning.</p>
<p>My ability to read Russian is limited to pressing buttons on Google Translate &#8211; and assuming that the occasional strange word is the wrong  literal translation of a word that can be used in many circumstances. Within the bounds that limitation, it seemed that most of the text made sense &#8211; but I hope readers following will be duly cautious.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Kumar</title>
		<link>http://jonathanlewis.wordpress.com/2010/12/17/system-statistics-2/#comment-38806</link>
		<dc:creator><![CDATA[Kumar]]></dc:creator>
		<pubDate>Tue, 21 Dec 2010 03:09:56 +0000</pubDate>
		<guid isPermaLink="false">http://jonathanlewis.wordpress.com/?p=4949#comment-38806</guid>
		<description><![CDATA[Hi Jonathan
As per note 368252.1, system statistics are not recommended for Oracle Applications and the reason quoted and I paste it  here is this.
But I am not sure if this is the right recommendation because the reason they quote is valid for any application not just oracle applications.  I am going to raise an SR with Oracle but wanted to understand what are your thoughts on this (Also the note was updated on Aug 2010 which makes me think they still think this is valid recommendation )
&lt;ul&gt;
&lt;em&gt;7) Should dbms_stats.gather_system_stats be used with E-Business 11i or R12?

The use of dbms_stats.gather_system_stats to collect system stats is not recommended with Apps E-Business Suite 11i or R12.

Not collecting system stats will make the CBO (in DB 10g onwards) to use what are called &quot;noworkload_stats&quot;, which have quite sensible defaults.

The main reason for this recommendation is the desire for stability and the potential problems related to large plan variances from system to system and even within one system, which would depend on the workload when these stats were gathered&lt;/em&gt;
&lt;/ul&gt;
Thank you
Kumar]]></description>
		<content:encoded><![CDATA[<p>Hi Jonathan<br />
As per note 368252.1, system statistics are not recommended for Oracle Applications and the reason quoted and I paste it  here is this.<br />
But I am not sure if this is the right recommendation because the reason they quote is valid for any application not just oracle applications.  I am going to raise an SR with Oracle but wanted to understand what are your thoughts on this (Also the note was updated on Aug 2010 which makes me think they still think this is valid recommendation )</p>
<ul>
<em>7) Should dbms_stats.gather_system_stats be used with E-Business 11i or R12?</p>
<p>The use of dbms_stats.gather_system_stats to collect system stats is not recommended with Apps E-Business Suite 11i or R12.</p>
<p>Not collecting system stats will make the CBO (in DB 10g onwards) to use what are called &#8220;noworkload_stats&#8221;, which have quite sensible defaults.</p>
<p>The main reason for this recommendation is the desire for stability and the potential problems related to large plan variances from system to system and even within one system, which would depend on the workload when these stats were gathered</em>
</ul>
<p>Thank you<br />
Kumar</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: joel garry</title>
		<link>http://jonathanlewis.wordpress.com/2010/12/17/system-statistics-2/#comment-38805</link>
		<dc:creator><![CDATA[joel garry]]></dc:creator>
		<pubDate>Mon, 20 Dec 2010 23:11:42 +0000</pubDate>
		<guid isPermaLink="false">http://jonathanlewis.wordpress.com/?p=4949#comment-38805</guid>
		<description><![CDATA[&lt;a href=&quot;http://en.wikipedia.org/wiki/Mister_Peabody&quot; rel=&quot;nofollow&quot;&gt;Peabody here&lt;/a&gt;.  http://web.archive.org/web/20040624011940/otn.oracle.com/pub/articles/lewis_cbo.html

Which just goes to show, you can lead the CBO to order, but you can&#039;t make it think.]]></description>
		<content:encoded><![CDATA[<p><a href="http://en.wikipedia.org/wiki/Mister_Peabody" rel="nofollow">Peabody here</a>.  <a href="http://web.archive.org/web/20040624011940/otn.oracle.com/pub/articles/lewis_cbo.html" rel="nofollow">http://web.archive.org/web/20040624011940/otn.oracle.com/pub/articles/lewis_cbo.html</a></p>
<p>Which just goes to show, you can lead the CBO to order, but you can&#8217;t make it think.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: John Hurley</title>
		<link>http://jonathanlewis.wordpress.com/2010/12/17/system-statistics-2/#comment-38778</link>
		<dc:creator><![CDATA[John Hurley]]></dc:creator>
		<pubDate>Fri, 17 Dec 2010 20:32:48 +0000</pubDate>
		<guid isPermaLink="false">http://jonathanlewis.wordpress.com/?p=4949#comment-38778</guid>
		<description><![CDATA[Should we now be hoping for an updated current posting soon on that topic with your most current thinking?  An early XMAS present or Happy New Year item?

Keeping my fingers crossed!]]></description>
		<content:encoded><![CDATA[<p>Should we now be hoping for an updated current posting soon on that topic with your most current thinking?  An early XMAS present or Happy New Year item?</p>
<p>Keeping my fingers crossed!</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: al0</title>
		<link>http://jonathanlewis.wordpress.com/2010/12/17/system-statistics-2/#comment-38777</link>
		<dc:creator><![CDATA[al0]]></dc:creator>
		<pubDate>Fri, 17 Dec 2010 19:55:24 +0000</pubDate>
		<guid isPermaLink="false">http://jonathanlewis.wordpress.com/?p=4949#comment-38777</guid>
		<description><![CDATA[Hi, 

better ask to remove that Russian translation (http://www.oracle.com/global/ru/oramag/march2005/admin_sys_stat.html), it is enormously bad.

Oleksandr]]></description>
		<content:encoded><![CDATA[<p>Hi, </p>
<p>better ask to remove that Russian translation (<a href="http://www.oracle.com/global/ru/oramag/march2005/admin_sys_stat.html" rel="nofollow">http://www.oracle.com/global/ru/oramag/march2005/admin_sys_stat.html</a>), it is enormously bad.</p>
<p>Oleksandr</p>
]]></content:encoded>
	</item>
</channel>
</rss>
