You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@jackrabbit.apache.org by ChadDavis <ch...@gmail.com> on 2010/01/18 20:30:27 UTC

jcr2spi spi2dav performance issue

I'm load testing my application.  The application is a web application
that connects to a remote Jackrabbit server ( which is the jackrabbit
webapp deployed in a Jetty instance ).  The connection is using
jcr2spi + spi2davex.  I'm on jackrabbit 2.0-beta3.

Our load tests show that the client side ( my web application ) is
creating new TCP connections at a very high rate.  Most of these TCP
connections go idle ( TIME_WAIT versus ESTABLISHED ).  On the
repository side, there is a set number of TCP connections that appear
to be re-used -- this is Jetty's code, I belive.

Why does the client network stack of Jackrabbit create such a
proliferation of TCP connections.  I'm pretty sure this is causing a
huge performance hit; the network bandwith is modest and well with in
our means.  Isn't there some reuse of TCP connections in this network
stack?  Perhaps I can do some configuration?

Re: jcr2spi spi2dav performance issue

Posted by ChadDavis <ch...@gmail.com>.
Ping.   Does anyone know anything about this?

I looked into the code a bit and it seems like that it might be a case
of the HttpClient api usage.  I'd be willing to invest some time into
this issue, provided that some one could give me some pointers.


On Mon, Jan 18, 2010 at 12:30 PM, ChadDavis <ch...@gmail.com> wrote:
> I'm load testing my application.  The application is a web application
> that connects to a remote Jackrabbit server ( which is the jackrabbit
> webapp deployed in a Jetty instance ).  The connection is using
> jcr2spi + spi2davex.  I'm on jackrabbit 2.0-beta3.
>
> Our load tests show that the client side ( my web application ) is
> creating new TCP connections at a very high rate.  Most of these TCP
> connections go idle ( TIME_WAIT versus ESTABLISHED ).  On the
> repository side, there is a set number of TCP connections that appear
> to be re-used -- this is Jetty's code, I belive.
>
> Why does the client network stack of Jackrabbit create such a
> proliferation of TCP connections.  I'm pretty sure this is causing a
> huge performance hit; the network bandwith is modest and well with in
> our means.  Isn't there some reuse of TCP connections in this network
> stack?  Perhaps I can do some configuration?
>