<?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/"
		>
<channel>
	<title>Comments on: I/O schedulers seriously revisited</title>
	<atom:link href="http://mituzas.lt/2008/02/05/linux-io-schedulers/feed/" rel="self" type="application/rss+xml" />
	<link>http://mituzas.lt/2008/02/05/linux-io-schedulers/</link>
	<description></description>
	<lastBuildDate>Fri, 30 Jul 2010 21:52:54 +0000</lastBuildDate>
	<sy:updatePeriod>hourly</sy:updatePeriod>
	<sy:updateFrequency>1</sy:updateFrequency>
	<generator>http://wordpress.org/?v=3.1-alpha</generator>
	<item>
		<title>By: Peter Zaitsev</title>
		<link>http://mituzas.lt/2008/02/05/linux-io-schedulers/comment-page-1/#comment-188795</link>
		<dc:creator>Peter Zaitsev</dc:creator>
		<pubDate>Sat, 31 Jan 2009 06:55:14 +0000</pubDate>
		<guid isPermaLink="false">http://dammit.lt/2008/02/05/linux-io-schedulers/#comment-188795</guid>
		<description>One interesting thing to note - the schedules keep their names but they are constantly tuned.  The behavior I observed for &quot;as&quot; vs &quot;deadline&quot; few years ago (well before cfq was born) is not the same as it is now.</description>
		<content:encoded><![CDATA[<p>One interesting thing to note &#8211; the schedules keep their names but they are constantly tuned.  The behavior I observed for &#8220;as&#8221; vs &#8220;deadline&#8221; few years ago (well before cfq was born) is not the same as it is now.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Domas Mituzas</title>
		<link>http://mituzas.lt/2008/02/05/linux-io-schedulers/comment-page-1/#comment-188793</link>
		<dc:creator>Domas Mituzas</dc:creator>
		<pubDate>Sat, 31 Jan 2009 00:44:23 +0000</pubDate>
		<guid isPermaLink="false">http://dammit.lt/2008/02/05/linux-io-schedulers/#comment-188793</guid>
		<description>William, &#039;just&#039; ;-)</description>
		<content:encoded><![CDATA[<p>William, &#8216;just&#8217; ;-)</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: William Newton</title>
		<link>http://mituzas.lt/2008/02/05/linux-io-schedulers/comment-page-1/#comment-188792</link>
		<dc:creator>William Newton</dc:creator>
		<pubDate>Fri, 30 Jan 2009 21:12:02 +0000</pubDate>
		<guid isPermaLink="false">http://dammit.lt/2008/02/05/linux-io-schedulers/#comment-188792</guid>
		<description>&quot;The worst part is that Deadline is enabled by default just on community distributions, like Ubuntu Server (though Ubuntu desktop kernel has CFQ by default).&quot;

What? From your test results it looks like deadline is much better.  If its the default on servers, thats a good thing. I think. Please elaberate on what you meant.</description>
		<content:encoded><![CDATA[<p>&#8220;The worst part is that Deadline is enabled by default just on community distributions, like Ubuntu Server (though Ubuntu desktop kernel has CFQ by default).&#8221;</p>
<p>What? From your test results it looks like deadline is much better.  If its the default on servers, thats a good thing. I think. Please elaberate on what you meant.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Domas Mituzas</title>
		<link>http://mituzas.lt/2008/02/05/linux-io-schedulers/comment-page-1/#comment-152145</link>
		<dc:creator>Domas Mituzas</dc:creator>
		<pubDate>Mon, 26 May 2008 20:02:42 +0000</pubDate>
		<guid isPermaLink="false">http://dammit.lt/2008/02/05/linux-io-schedulers/#comment-152145</guid>
		<description>Nat, 
there were only writes - reads generally are served from buffer pool.
There were no other I/O requests.

ionice doesn&#039;t help that much - as it is scheduler dependent, and also - it is just one process.

BTW, the box is 4-core, not 8-core.</description>
		<content:encoded><![CDATA[<p>Nat,<br />
there were only writes &#8211; reads generally are served from buffer pool.<br />
There were no other I/O requests.</p>
<p>ionice doesn&#8217;t help that much &#8211; as it is scheduler dependent, and also &#8211; it is just one process.</p>
<p>BTW, the box is 4-core, not 8-core.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Nat</title>
		<link>http://mituzas.lt/2008/02/05/linux-io-schedulers/comment-page-1/#comment-127641</link>
		<dc:creator>Nat</dc:creator>
		<pubDate>Sun, 23 Mar 2008 15:15:13 +0000</pubDate>
		<guid isPermaLink="false">http://dammit.lt/2008/02/05/linux-io-schedulers/#comment-127641</guid>
		<description>&gt; I created a very simple table and started spamming rows

Did you only write or were there a few reads?
Were there other I/O requests (non submitted by Mysql)?
Did you use &#039;ionice&#039; (Debian: &#039;schedutils&#039; package) in order to set the MySQL process in an appropriate (favored!) I/O scheduling class?
Example: &quot;ionice -c2 -n0 /usr/bin/mysqld_safe ...&quot;
(the &quot;realtime&quot; class seems not very efficient, see my website)</description>
		<content:encoded><![CDATA[<p>&gt; I created a very simple table and started spamming rows</p>
<p>Did you only write or were there a few reads?<br />
Were there other I/O requests (non submitted by Mysql)?<br />
Did you use &#8216;ionice&#8217; (Debian: &#8216;schedutils&#8217; package) in order to set the MySQL process in an appropriate (favored!) I/O scheduling class?<br />
Example: &#8220;ionice -c2 -n0 /usr/bin/mysqld_safe &#8230;&#8221;<br />
(the &#8220;realtime&#8221; class seems not very efficient, see my website)</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: john allspaw</title>
		<link>http://mituzas.lt/2008/02/05/linux-io-schedulers/comment-page-1/#comment-108537</link>
		<dc:creator>john allspaw</dc:creator>
		<pubDate>Mon, 11 Feb 2008 21:11:10 +0000</pubDate>
		<guid isPermaLink="false">http://dammit.lt/2008/02/05/linux-io-schedulers/#comment-108537</guid>
		<description>IIRC, only some newer kernels allow for on-the-fly IO scheduler changes.  

We also have our squids set with deadline as well as our DBs.</description>
		<content:encoded><![CDATA[<p>IIRC, only some newer kernels allow for on-the-fly IO scheduler changes.  </p>
<p>We also have our squids set with deadline as well as our DBs.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Domas Mituzas</title>
		<link>http://mituzas.lt/2008/02/05/linux-io-schedulers/comment-page-1/#comment-106358</link>
		<dc:creator>Domas Mituzas</dc:creator>
		<pubDate>Wed, 06 Feb 2008 07:52:53 +0000</pubDate>
		<guid isPermaLink="false">http://dammit.lt/2008/02/05/linux-io-schedulers/#comment-106358</guid>
		<description>pardon my lingo, innodb-flush-log-at-trx-commit :)</description>
		<content:encoded><![CDATA[<p>pardon my lingo, innodb-flush-log-at-trx-commit :)</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Xing</title>
		<link>http://mituzas.lt/2008/02/05/linux-io-schedulers/comment-page-1/#comment-106316</link>
		<dc:creator>Xing</dc:creator>
		<pubDate>Wed, 06 Feb 2008 04:42:21 +0000</pubDate>
		<guid isPermaLink="false">http://dammit.lt/2008/02/05/linux-io-schedulers/#comment-106316</guid>
		<description>What is flush=0? Where is this set?</description>
		<content:encoded><![CDATA[<p>What is flush=0? Where is this set?</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Partha Dutta</title>
		<link>http://mituzas.lt/2008/02/05/linux-io-schedulers/comment-page-1/#comment-106232</link>
		<dc:creator>Partha Dutta</dc:creator>
		<pubDate>Tue, 05 Feb 2008 23:39:02 +0000</pubDate>
		<guid isPermaLink="false">http://dammit.lt/2008/02/05/linux-io-schedulers/#comment-106232</guid>
		<description>Very curious about the flush=0 flag. Is this a boot parameter or an option for the deadline scheduler?

Thanks, Partha</description>
		<content:encoded><![CDATA[<p>Very curious about the flush=0 flag. Is this a boot parameter or an option for the deadline scheduler?</p>
<p>Thanks, Partha</p>
]]></content:encoded>
	</item>
</channel>
</rss>
