Thank you everyone for the responses/suggestions. I did ensure that those were the same during the tests, but had removed that from the server.xml before sending. Both servers had negotiated the equivalent of TLS_RSA_WITH_AES_128_CBC_SHA with the client during testing.
Also, i've verified the JVM is in server mode, and I'm currently using 1.6.0_14. At times i can hit increased speeds of 10.8MB/s for IIS (about as fast as this server's network card can handle), with tomcat running close to 7.8 MB/s. (I'm just thinking out loud here) Assuming this is a normal difference with tomcat running at 72% the speed of IIS, and if the system was congested to the point of slowing down the downloads from the servers, then I would expect tomcat to run at 2.5 MB/s if IIS was serving files at 3.5 MB/s. It however looks like tomcat is serving data at a consistent 3.0 MB/s slower than IIS (not percentage based), which is why i saw 350 KB/s against the 3.5 MB/s from IIS? Anyway, I will try the APR connector and let you know. Thanks again for your comments! On Thu, Jan 10, 2013 at 8:31 AM, David kerber <dcker...@verizon.net> wrote: > On 1/10/2013 8:56 AM, Linoma DevTeam wrote: > >> Hi everyone, >> >> I'm running some comparison tests with tomcat 6.0.35 and IIS running in >> parallel on Windows Server 2008 R2. Now I would expect Tomcat to be >> somewhat slower, given the extra JVM layer, but in some situations, i'm >> seeing differences that are tough to swallow. >> > > What JVM are you running under? Is it running the client or server > version? I've found huge performance improvements in some kind of > operations under a server JVM compared to client ones. > > > D > > > >> Downloads >> IIS ~3.7 MB/s >> Tomcat ~350 KB/s >> >> Test Details: >> I placed a ~500MB file in the document root of the web app on tomcat and >> set up an HTTPS connector. Then I set up IIS with the same file and an >> HTTPS listener. I configured the cipher suite in tomcat to be the same >> one >> that was negotiated between IIS and my Chrome browser. Finally, I set the >> JVM max memory to 1024MB with a min of 900MB to reduce the impact of the >> GC >> and the memory allocation. >> >> I'm using HTTP/1.1 connectors with pretty standard configuration: >> >> <Server port="9005" shutdown="SHUTDOWN"> >> >> <Service name="admin"> >> <Connector port="9080" /> >> >> <Connector port="9443" protocol="HTTP/1.1" >> SSLEnabled="true" enableLookups="false" disableUploadTimeout="true" >> scheme="https" secure="true" clientAuth="false" sslProtocol="TLS" >> algorithm="SunX509" keystoreFile="C:\temp\sample_**keystore.jks" >> keystorePass="password" keyAlias="sample-key" keystoreType="JKS" >> truststoreFile="C:\temp\**sample_truststore.jks" >> truststorePass="password" >> truststoreType="JKS" /> >> >> <Engine name="admin" defaultHost="localhost"> >> <Host name="localhost" appBase="webapps" >> errorReportValveClass="com.**company.**CustomErrorReportValve"> >> <Context path="/sample" docBase="C:\temp\application\**WebRoot" >> reloadable="false"> >> <Loader delegate="true" /> >> </Context> >> </Host> >> </Engine> >> </Service> >> </Server> >> >> So, I extend the question of, why would tomcat only be able to reach 10% >> of >> the speed IIS is able to server when running parallel tests? Any >> suggestions on configurations that I could adjust on Tomcat, the JVM, or >> operating system that improve that download speed? >> >> Thanks in advance! >> >> > > ------------------------------**------------------------------**--------- > To unsubscribe, e-mail: > users-unsubscribe@tomcat.**apache.org<users-unsubscr...@tomcat.apache.org> > For additional commands, e-mail: users-h...@tomcat.apache.org > >
