<?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: ASMLIB Performance vs Udev</title>
	<atom:link href="http://www.ardentperf.com/2008/10/08/asmlib-performance-vs-udev/feed/" rel="self" type="application/rss+xml" />
	<link>http://www.ardentperf.com/2008/10/08/asmlib-performance-vs-udev/</link>
	<description>Jeremy's Oracle Resources and Ramblings</description>
	<lastBuildDate>Wed, 03 Mar 2010 03:35:43 +0000</lastBuildDate>
	<generator>http://wordpress.org/?v=2.9.1</generator>
	<sy:updatePeriod>hourly</sy:updatePeriod>
	<sy:updateFrequency>1</sy:updateFrequency>
		<item>
		<title>By: Jeremy</title>
		<link>http://www.ardentperf.com/2008/10/08/asmlib-performance-vs-udev/comment-page-1/#comment-1529</link>
		<dc:creator>Jeremy</dc:creator>
		<pubDate>Thu, 12 Mar 2009 18:24:26 +0000</pubDate>
		<guid isPermaLink="false">http://www.ardentperf.com/?p=615#comment-1529</guid>
		<description>One more caveat. We&#039;ve pushed ahead with using udev here. Every time a new hardware model is introduced we have to figure out new udev rules. (Multipath, SCSI, SSD, HP cciss hardware RAID...) ASMLIB probably would have been a bit easier in this regard. (I&#039;m not certain off-hand, but I think ASMLIB works with any block device.)

Basically my current thinking is that you should use udev if you have Linux guys around to take care of it, ASMLIB if the DBAs are going to be tending the device permissions. Guess it kinda relates to that whole DBA2.0 thing that was recently making the rounds.</description>
		<content:encoded><![CDATA[<p>One more caveat. We&#8217;ve pushed ahead with using udev here. Every time a new hardware model is introduced we have to figure out new udev rules. (Multipath, SCSI, SSD, HP cciss hardware RAID&#8230;) ASMLIB probably would have been a bit easier in this regard. (I&#8217;m not certain off-hand, but I think ASMLIB works with any block device.)</p>
<p>Basically my current thinking is that you should use udev if you have Linux guys around to take care of it, ASMLIB if the DBAs are going to be tending the device permissions. Guess it kinda relates to that whole DBA2.0 thing that was recently making the rounds.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Jeremy</title>
		<link>http://www.ardentperf.com/2008/10/08/asmlib-performance-vs-udev/comment-page-1/#comment-1528</link>
		<dc:creator>Jeremy</dc:creator>
		<pubDate>Thu, 05 Mar 2009 02:06:34 +0000</pubDate>
		<guid isPermaLink="false">http://www.ardentperf.com/?p=615#comment-1528</guid>
		<description>One recent update... I haven&#039;t verified this, but:

I heard that with regular block devices, after dropping or offline-ing a device it might still be help open by other rdbms processes besides the one that did the drop. I&#039;m not sure if or how this might cause problems. (Do you need all the handles cleanly closed for OS operations like detaching the LUN or renaming the device?) I imagine that ASMLIB would not have this problem since each server process uses a single ASMLIB handle instead of direct device handles.

Have yet to verify the details. :)</description>
		<content:encoded><![CDATA[<p>One recent update&#8230; I haven&#8217;t verified this, but:</p>
<p>I heard that with regular block devices, after dropping or offline-ing a device it might still be help open by other rdbms processes besides the one that did the drop. I&#8217;m not sure if or how this might cause problems. (Do you need all the handles cleanly closed for OS operations like detaching the LUN or renaming the device?) I imagine that ASMLIB would not have this problem since each server process uses a single ASMLIB handle instead of direct device handles.</p>
<p>Have yet to verify the details. :)</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Ludovico Caldara</title>
		<link>http://www.ardentperf.com/2008/10/08/asmlib-performance-vs-udev/comment-page-1/#comment-1425</link>
		<dc:creator>Ludovico Caldara</dc:creator>
		<pubDate>Wed, 10 Dec 2008 08:55:03 +0000</pubDate>
		<guid isPermaLink="false">http://www.ardentperf.com/?p=615#comment-1425</guid>
		<description>Wow, I didn&#039;t figured out the possibility of using Udev with ASM before reading your post. That&#039;s great.
I&#039;ll follow your blog with entusiasm!</description>
		<content:encoded><![CDATA[<p>Wow, I didn&#8217;t figured out the possibility of using Udev with ASM before reading your post. That&#8217;s great.<br />
I&#8217;ll follow your blog with entusiasm!</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Jeremy</title>
		<link>http://www.ardentperf.com/2008/10/08/asmlib-performance-vs-udev/comment-page-1/#comment-1360</link>
		<dc:creator>Jeremy</dc:creator>
		<pubDate>Tue, 14 Oct 2008 15:05:50 +0000</pubDate>
		<guid isPermaLink="false">http://www.ardentperf.com/?p=615#comment-1360</guid>
		<description>Dan - thanks for stopping by! Fixed the link, thanks for pointing it out... I have it correct on my &quot;Links&quot; page but somehow I got the last name mixed up here.

Fritz and Jason - thanks for the comments. I agree that udev is a bit difficult; it&#039;s not exactly obvious and the documentation is very poor. But once you figure it out, it&#039;s quite powerful and flexible.</description>
		<content:encoded><![CDATA[<p>Dan &#8211; thanks for stopping by! Fixed the link, thanks for pointing it out&#8230; I have it correct on my &#8220;Links&#8221; page but somehow I got the last name mixed up here.</p>
<p>Fritz and Jason &#8211; thanks for the comments. I agree that udev is a bit difficult; it&#8217;s not exactly obvious and the documentation is very poor. But once you figure it out, it&#8217;s quite powerful and flexible.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Dan Norris</title>
		<link>http://www.ardentperf.com/2008/10/08/asmlib-performance-vs-udev/comment-page-1/#comment-1359</link>
		<dc:creator>Dan Norris</dc:creator>
		<pubDate>Mon, 13 Oct 2008 18:24:40 +0000</pubDate>
		<guid isPermaLink="false">http://www.ardentperf.com/?p=615#comment-1359</guid>
		<description>One small update: s/gorman/hall/.

Tim Hall is widely known as &quot;Tim...&quot;. Though I think he may also be the original and only case of an illness called &quot;Oracle Open World Syndrome (OOWS)&quot; which seems to affect him each year he attends (at least two years running now). 

Thanks for the blog. Hope to see more like this one coming soon!</description>
		<content:encoded><![CDATA[<p>One small update: s/gorman/hall/.</p>
<p>Tim Hall is widely known as &#8220;Tim&#8230;&#8221;. Though I think he may also be the original and only case of an illness called &#8220;Oracle Open World Syndrome (OOWS)&#8221; which seems to affect him each year he attends (at least two years running now). </p>
<p>Thanks for the blog. Hope to see more like this one coming soon!</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: jason arneil</title>
		<link>http://www.ardentperf.com/2008/10/08/asmlib-performance-vs-udev/comment-page-1/#comment-1358</link>
		<dc:creator>jason arneil</dc:creator>
		<pubDate>Thu, 09 Oct 2008 09:15:59 +0000</pubDate>
		<guid isPermaLink="false">http://www.ardentperf.com/?p=615#comment-1358</guid>
		<description>Hello,

Like Frits, I have also presented on ASM. Though I have a more benign attitude to ASM, but then I found udev to be quite impenetrable to use.

I have Oracle employees express the opinion the ASMLib enables global opening of devices, so 1 file handler per device which is shared amongst processes rather than each process having to do this.

I really have to test this so you&#039;d treat this with a large pinch of salt until there is more concrete evidence.

Of course, whether reducing file handlers, and marginal improvement in context switches (thus reducing cpu) is really a benefit you will notice is a different question.

jason.</description>
		<content:encoded><![CDATA[<p>Hello,</p>
<p>Like Frits, I have also presented on ASM. Though I have a more benign attitude to ASM, but then I found udev to be quite impenetrable to use.</p>
<p>I have Oracle employees express the opinion the ASMLib enables global opening of devices, so 1 file handler per device which is shared amongst processes rather than each process having to do this.</p>
<p>I really have to test this so you&#8217;d treat this with a large pinch of salt until there is more concrete evidence.</p>
<p>Of course, whether reducing file handlers, and marginal improvement in context switches (thus reducing cpu) is really a benefit you will notice is a different question.</p>
<p>jason.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Frits Hoogland</title>
		<link>http://www.ardentperf.com/2008/10/08/asmlib-performance-vs-udev/comment-page-1/#comment-1357</link>
		<dc:creator>Frits Hoogland</dc:creator>
		<pubDate>Wed, 08 Oct 2008 12:46:08 +0000</pubDate>
		<guid isPermaLink="false">http://www.ardentperf.com/?p=615#comment-1357</guid>
		<description>I&#039;ve done the investigation and came to the same conclusion. It&#039;s also part of my ASM presentation I&#039;ve done on openworld.

There are more things you have to be aware of when using asmlib:
-the asmlib library is placed in /opt/oracle/extapi/32/asm/orcl/1/libasm.so. No problem, but if you are not aware you have a critical library placed there, it&#039;s possible that someone (a sysadmin) cleanes this up.
-asmlib adds a kernel dependency. No problem, but you have to be aware you have to upgrade the asmlib kernel package together with upgrading your kernel. 
-what sense does a seperate asmlib name make for a disk? it means you have a device name (which can differ), a ASM diskname, and a asmlib name. seems redundant to me. also, when udev is setup to give the disks for oracle the correct permissions (and the asm discovery string is setup right), ASM reads the diskheaders and recognises them, no matter if disk /dev/sdd1 was /dev/sdb1 before the reboot.</description>
		<content:encoded><![CDATA[<p>I&#8217;ve done the investigation and came to the same conclusion. It&#8217;s also part of my ASM presentation I&#8217;ve done on openworld.</p>
<p>There are more things you have to be aware of when using asmlib:<br />
-the asmlib library is placed in /opt/oracle/extapi/32/asm/orcl/1/libasm.so. No problem, but if you are not aware you have a critical library placed there, it&#8217;s possible that someone (a sysadmin) cleanes this up.<br />
-asmlib adds a kernel dependency. No problem, but you have to be aware you have to upgrade the asmlib kernel package together with upgrading your kernel.<br />
-what sense does a seperate asmlib name make for a disk? it means you have a device name (which can differ), a ASM diskname, and a asmlib name. seems redundant to me. also, when udev is setup to give the disks for oracle the correct permissions (and the asm discovery string is setup right), ASM reads the diskheaders and recognises them, no matter if disk /dev/sdd1 was /dev/sdb1 before the reboot.</p>
]]></content:encoded>
	</item>
</channel>
</rss>
