benchmarks?
4 views
Skip to first unread message
Brian Hammond
Jun 14, 2009, 4:29:15 PM6/14/09
to Hemlock
I do not see anywhere benchmarks of a real-world deployment.
If you keep a connection open, your server(s) are going to be limited
to some number (say 50,000) of simultaneous connections.
What is the server side ultimately based on? libev, libevent, kqueue/
epoll/etc.?
If you keep a connection open, your server(s) are going to be limited
to some number (say 50,000) of simultaneous connections.
What is the server side ultimately based on? libev, libevent, kqueue/
epoll/etc.?
Cameron Price
Jun 17, 2009, 5:33:37 PM6/17/09
to Hemlock
Brian, it kinda depends. The server side of XMPP is entirely
dependent on your XMPP server. In our case we're using ejabberd,
which >claims< to scale up massively, and is written in erlang, which
is usually good about such things. However, I can't say for sure what
the scaling pain points will be as none of our hemlock apps have yet
his anything like large numbers.
That said, serving XMPP in general, and on a large scale, is a fairly
solved problem, and I can't see a reason why a hemlock app would
exhibit different performance characteristics to a similar number of
chat clients.
-c
On Jun 14, 4:29 pm, Brian Hammond <or.else.it.gets.the.h...@gmail.com>
wrote:
dependent on your XMPP server. In our case we're using ejabberd,
which >claims< to scale up massively, and is written in erlang, which
is usually good about such things. However, I can't say for sure what
the scaling pain points will be as none of our hemlock apps have yet
his anything like large numbers.
That said, serving XMPP in general, and on a large scale, is a fairly
solved problem, and I can't see a reason why a hemlock app would
exhibit different performance characteristics to a similar number of
chat clients.
-c
On Jun 14, 4:29 pm, Brian Hammond <or.else.it.gets.the.h...@gmail.com>
wrote:
Reply all
Reply to author
Forward
0 new messages