How heavy is TIMP in terms of CPU load ?

The CPU load does not depends on the number of client but on the number of messages routed.

Usually, also the roster size affects performance. At tipic.com we run TIMP with a mean many hundreds concurrent users in busy hours so DEFINITELY the problem is not in the number of connected clients.

TIMP uses a kind of "traffic shaping" to limit the number of messages per second a client can send; if a client exceeds that limit it happens to be "karmed" in the O.S. version of jabber (which lead to
messages lost) or simply disconnected in TIMP.

If you have developed a server side plug-in plug-in you have to check if you are sending too many messages; in this case the client gets disconnected and then it sudden reconnects in a neverending loop which in fact might lead to 100% CPU usage.

You can easly check the number of messages that are circulating in the system either by using TIMP console / Server log (remember to press on the "Update Log" button) or just looking in the text file "C:Documents and settingsTW32JS_USERApplication DataTipicMainAte.log" where you can find some rows like "Router: xxx messages successiful routed". The server writes such a line about every 30 seconds, so subtracting the value found in the last two entries you get half the messages per minute TIMP is routing.

Another cause of 100 CPU would be a kind of Misconfiguration in which TIMP sends a message to the s2s component and this then re-send it back to TIMP in a neverending loop.