Spikes in TCP Networking
 

I am monitoring the transmission and receipt of messages over two linux machines across a point to point TCP network. The messages are sent at a constant rate (10milliseconds between messages), but at regular intervals (30seconds), a large spike in the time between a couple of messages (upto 80milliseconds) occurs.

With some messing about (using a more powerful machine, turning off the "cron" event etc), I have managed to get the tranmission rate to a constant 10m/s with no spikes, but the correlating received log still shows the regular spiking.

Are there any system/kernel events that I am unaware of that may be causing a regular slight 'pause'?

Any ideas/suggestions as to why I am getting these spikes would be much appreciated.