<?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: Solid State Devices (SSD) and journaling</title>
	<atom:link href="http://beyondabstraction.net/2009/03/02/solid-state-devices-ssd-and-journaling/feed/" rel="self" type="application/rss+xml" />
	<link>http://beyondabstraction.net/2009/03/02/solid-state-devices-ssd-and-journaling/</link>
	<description>Meanderings and Such...</description>
	<lastBuildDate>Sat, 12 Jun 2010 13:18:54 +0000</lastBuildDate>
	<sy:updatePeriod>hourly</sy:updatePeriod>
	<sy:updateFrequency>1</sy:updateFrequency>
	<generator>http://wordpress.org/?v=3.0</generator>
	<item>
		<title>By: spencer</title>
		<link>http://beyondabstraction.net/2009/03/02/solid-state-devices-ssd-and-journaling/comment-page-1/#comment-24906</link>
		<dc:creator>spencer</dc:creator>
		<pubDate>Mon, 16 Mar 2009 02:49:47 +0000</pubDate>
		<guid isPermaLink="false">http://beyondabstraction.net/?p=307#comment-24906</guid>
		<description>Thanks for elaborating Ted.  

Would an accurate &quot;short and sweet&quot; description be (?):

FAT/VFAT files and directories are less distributed than it&#039;s counterparts like ext234/NTFS.  Therefore FAT is less likely to suffer from the same write amplification problems.</description>
		<content:encoded><![CDATA[<p>Thanks for elaborating Ted.  </p>
<p>Would an accurate &#8220;short and sweet&#8221; description be (?):</p>
<p>FAT/VFAT files and directories are less distributed than it&#8217;s counterparts like ext234/NTFS.  Therefore FAT is less likely to suffer from the same write amplification problems.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Theodore Tso</title>
		<link>http://beyondabstraction.net/2009/03/02/solid-state-devices-ssd-and-journaling/comment-page-1/#comment-24270</link>
		<dc:creator>Theodore Tso</dc:creator>
		<pubDate>Mon, 02 Mar 2009 13:26:48 +0000</pubDate>
		<guid isPermaLink="false">http://beyondabstraction.net/?p=307#comment-24270</guid>
		<description>&lt;i&gt;The bottom-line as far as wear and tear: vfat is basically the same as a journaled FS from a wear-leveling standpoint, both are worst case.&lt;/i&gt;

Well..... not necessarily.  You&#039;re overstating the case I made.   Many flash drives are only doing wear leveling at the erase block level, and so they suffer from the write amplification effect --- an update of a single 4k block in an 128k region may still require the update and rewrite on the entire 128k erase block.

VFAT is better suited for &#039;dumb&#039; first generation flash drives because if you are writing a single large file, you only need to update one, maybe two erase block&#039;s in the FAT area, plus the space for the large file, plus the directory entry.    So for digital cameras, and for moving files around, thumb drives and CF/SD cards work just fine.

Both ext2 and ext3 write more data to different parts of the disk.  For large files, what they write will likely be more efficient than FAT in terms of number of blocks modified on disk (the FAT table is astonishingly inefficient, since you have to modify 4 bytes for every 8-32k of data blocks that are modified), but the writes are scattered across the disk.   So for Flash drives that aren&#039;t doing wear-leveling at the 512 byte or 4k block level, but rather at the 128k erase block level, the write amplification effect will cause the flash to be worn down much more rapidly than you might expect simply looking at the number of blocks written.   This isn&#039;t an issue for the Intel SSD&#039;s, but it almost certainly is an issue for the $9.95 thumb drive for sale at grocery store hanging on the rack near the checkout counter next to the chewing gum.  :-)

And before you ask, NTFS has similar issue for a large file (say, a raw camera image), it probably requires fewer metadata block writes than FAT, but they aren&#039;t located close together, which means there will be a similar write amplification issue.</description>
		<content:encoded><![CDATA[<p><i>The bottom-line as far as wear and tear: vfat is basically the same as a journaled FS from a wear-leveling standpoint, both are worst case.</i></p>
<p>Well&#8230;.. not necessarily.  You&#8217;re overstating the case I made.   Many flash drives are only doing wear leveling at the erase block level, and so they suffer from the write amplification effect &#8212; an update of a single 4k block in an 128k region may still require the update and rewrite on the entire 128k erase block.</p>
<p>VFAT is better suited for &#8216;dumb&#8217; first generation flash drives because if you are writing a single large file, you only need to update one, maybe two erase block&#8217;s in the FAT area, plus the space for the large file, plus the directory entry.    So for digital cameras, and for moving files around, thumb drives and CF/SD cards work just fine.</p>
<p>Both ext2 and ext3 write more data to different parts of the disk.  For large files, what they write will likely be more efficient than FAT in terms of number of blocks modified on disk (the FAT table is astonishingly inefficient, since you have to modify 4 bytes for every 8-32k of data blocks that are modified), but the writes are scattered across the disk.   So for Flash drives that aren&#8217;t doing wear-leveling at the 512 byte or 4k block level, but rather at the 128k erase block level, the write amplification effect will cause the flash to be worn down much more rapidly than you might expect simply looking at the number of blocks written.   This isn&#8217;t an issue for the Intel SSD&#8217;s, but it almost certainly is an issue for the $9.95 thumb drive for sale at grocery store hanging on the rack near the checkout counter next to the chewing gum.  <img src='http://beyondabstraction.net/wp-includes/images/smilies/icon_smile.gif' alt=':-)' class='wp-smiley' /> </p>
<p>And before you ask, NTFS has similar issue for a large file (say, a raw camera image), it probably requires fewer metadata block writes than FAT, but they aren&#8217;t located close together, which means there will be a similar write amplification issue.</p>
]]></content:encoded>
	</item>
</channel>
</rss>
