<?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: Flash vs disk at DISKCON 2007</title>
	<atom:link href="http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/feed/" rel="self" type="application/rss+xml" />
	<link>http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/</link>
	<description>Data storage info &#38; analysis</description>
	<lastBuildDate>Fri, 19 Mar 2010 09:23:11 +0000</lastBuildDate>
	<generator>http://wordpress.org/?v=2.9.2</generator>
	<sy:updatePeriod>hourly</sy:updatePeriod>
	<sy:updateFrequency>1</sy:updateFrequency>
		<item>
		<title>By: Jim Thompson</title>
		<link>http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/comment-page-1/#comment-138389</link>
		<dc:creator>Jim Thompson</dc:creator>
		<pubDate>Sun, 28 Oct 2007 08:52:50 +0000</pubDate>
		<guid isPermaLink="false">http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/#comment-138389</guid>
		<description>You appear to be missing the contribution of SSDs to technologies like ZFS.  

First, you need a workload where the ZIL has an impact.

The anticipaged improvement comes from separating the ZIL workload, which is small, sequential iops with a high desire for low latency, from more random workloads. 
dedicated to the ZIL. Now think of how ZFS could use one or more SSDs to implement the ZIL, and specifically the types of access that work best with flash.

Now consider that hybrid drives (which you apparently hate) come with a &#039;flash partition&#039; which ZFS could use to spread the ZIL out over several drives...

and connect the dots.</description>
		<content:encoded><![CDATA[<p>You appear to be missing the contribution of SSDs to technologies like ZFS.  </p>
<p>First, you need a workload where the ZIL has an impact.</p>
<p>The anticipaged improvement comes from separating the ZIL workload, which is small, sequential iops with a high desire for low latency, from more random workloads.<br />
dedicated to the ZIL. Now think of how ZFS could use one or more SSDs to implement the ZIL, and specifically the types of access that work best with flash.</p>
<p>Now consider that hybrid drives (which you apparently hate) come with a &#8216;flash partition&#8217; which ZFS could use to spread the ZIL out over several drives&#8230;</p>
<p>and connect the dots.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Amir Hermelin</title>
		<link>http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/comment-page-1/#comment-136018</link>
		<dc:creator>Amir Hermelin</dc:creator>
		<pubDate>Mon, 22 Oct 2007 22:54:10 +0000</pubDate>
		<guid isPermaLink="false">http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/#comment-136018</guid>
		<description>Robin, in the post you said:
&quot;Rick had a larger point as well. The disk guys have an incredible story to tell and they aren’t telling it. I’ll be getting into some of that later this week.&quot;

Could you shed some light on that interesting disk guys&#039; story? Thanks!

- amir</description>
		<content:encoded><![CDATA[<p>Robin, in the post you said:<br />
&#8220;Rick had a larger point as well. The disk guys have an incredible story to tell and they aren’t telling it. I’ll be getting into some of that later this week.&#8221;</p>
<p>Could you shed some light on that interesting disk guys&#8217; story? Thanks!</p>
<p>- amir</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Bill Todd</title>
		<link>http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/comment-page-1/#comment-128377</link>
		<dc:creator>Bill Todd</dc:creator>
		<pubDate>Tue, 02 Oct 2007 10:30:41 +0000</pubDate>
		<guid isPermaLink="false">http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/#comment-128377</guid>
		<description>The number of concurrent streams still simply boils down to aggregate storage bandwidth, and for large sequential (i.e., streaming) accesses commodity disks can provide that at less than 1/10th the price that the least expensive flash can.  Even extreme hot spots where a large percentage of streams focus on a small percentage of files can be handled as long as the files are spread across the entire array, since the streams are staggered in the time dimension.  And for pathologically intense access cases, data caching in server RAM can perform the same service as flash without the cost of keeping *everything* in expensive high-speed storage.

- bill</description>
		<content:encoded><![CDATA[<p>The number of concurrent streams still simply boils down to aggregate storage bandwidth, and for large sequential (i.e., streaming) accesses commodity disks can provide that at less than 1/10th the price that the least expensive flash can.  Even extreme hot spots where a large percentage of streams focus on a small percentage of files can be handled as long as the files are spread across the entire array, since the streams are staggered in the time dimension.  And for pathologically intense access cases, data caching in server RAM can perform the same service as flash without the cost of keeping *everything* in expensive high-speed storage.</p>
<p>- bill</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Hans</title>
		<link>http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/comment-page-1/#comment-127305</link>
		<dc:creator>Hans</dc:creator>
		<pubDate>Fri, 28 Sep 2007 17:00:28 +0000</pubDate>
		<guid isPermaLink="false">http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/#comment-127305</guid>
		<description>It&#039;s the number of concurrent streams....</description>
		<content:encoded><![CDATA[<p>It&#8217;s the number of concurrent streams&#8230;.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Matt</title>
		<link>http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/comment-page-1/#comment-127236</link>
		<dc:creator>Matt</dc:creator>
		<pubDate>Fri, 28 Sep 2007 12:13:04 +0000</pubDate>
		<guid isPermaLink="false">http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/#comment-127236</guid>
		<description>Hans,

Why would you need flash for VOD (I take it you mean video on demand). Video is inherently disk friendly with its combination of very large files and large block sequential reads and writes. Outside of edge cases, flash is never going to be crazy useful in video.

Or are you doing something strange?</description>
		<content:encoded><![CDATA[<p>Hans,</p>
<p>Why would you need flash for VOD (I take it you mean video on demand). Video is inherently disk friendly with its combination of very large files and large block sequential reads and writes. Outside of edge cases, flash is never going to be crazy useful in video.</p>
<p>Or are you doing something strange?</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Hans</title>
		<link>http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/comment-page-1/#comment-126981</link>
		<dc:creator>Hans</dc:creator>
		<pubDate>Thu, 27 Sep 2007 16:42:09 +0000</pubDate>
		<guid isPermaLink="false">http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/#comment-126981</guid>
		<description>Have you looked at the Fusion-IO PCI-e based flash card introduced in DEMO?
I am looking at them for possible use in VOD applcations..</description>
		<content:encoded><![CDATA[<p>Have you looked at the Fusion-IO PCI-e based flash card introduced in DEMO?<br />
I am looking at them for possible use in VOD applcations..</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Barry Whyte</title>
		<link>http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/comment-page-1/#comment-125743</link>
		<dc:creator>Barry Whyte</dc:creator>
		<pubDate>Mon, 24 Sep 2007 16:10:48 +0000</pubDate>
		<guid isPermaLink="false">http://storagemojo.com/2007/09/24/flash-vs-disk-at-diskcon-2007/#comment-125743</guid>
		<description>Some comments on pure SSD over here Robin :

http://www.ibm.com/developerworks/blogs/page/storagevirtualization?entry=ssd_s_are_becoming_a</description>
		<content:encoded><![CDATA[<p>Some comments on pure SSD over here Robin :</p>
<p><a href="http://www.ibm.com/developerworks/blogs/page/storagevirtualization?entry=ssd_s_are_becoming_a" rel="nofollow">http://www.ibm.com/developerworks/blogs/page/storagevirtualization?entry=ssd_s_are_becoming_a</a></p>
]]></content:encoded>
	</item>
</channel>
</rss>
