<?xml version="1.0"?>
<rss version="2.0">
  <channel>
    <title>Advogato blog for jch</title>
    <link>http://www.advogato.org/person/jch/</link>
    <description>Advogato blog for jch</description>
    <language>en-us</language>
    <generator>mod_virgule</generator>
    <pubDate>Sun, 7 Sep 2008 03:23:55 GMT</pubDate>
    <item>
      <pubDate>Tue, 13 May 2008 11:58:34 GMT</pubDate>
      <title>13 May 2008</title>
      <link>http://www.advogato.org/person/jch/diary.html?start=0</link>
      <guid>http://www.advogato.org/person/jch/diary.html?start=0</guid>
      <description>&lt;p&gt;Gabriel Kerneis and myself have just finished compiling&#xD;
some &lt;a href="http://www.pps.jussieu.fr/~jch/software/cpc/bench/" &gt;latency&#xD;
statistics for web servers&lt;/a&gt;.  The results, assuming they&#xD;
are right, are rather surprising.&#xD;
&#xD;
&lt;p&gt; &lt;p&gt;As we expected, processes are horrible and event-driven&#xD;
code is very good.  The surprise is that kernel-threads,&#xD;
while unfair, are not as bad as we expected, and that&#xD;
user-space threads, if implemented right, are actually&#xD;
pretty good.&#xD;
&#xD;
&lt;p&gt; &lt;p&gt;Did anyone do anything similar?&#xD;
</description>
    </item>
  </channel>
</rss>
