We have one epi-installation for the editors; this server publishes the information and distributes it to three front slaves.

They all share the same database.

We have set up the cache invalidation using tcp/ip according to the guidelines and it works.

However we have noticed that the invalidation time varies up to five minutes for a random node.

This is not a acceptable time and the customer wants a instant page update on all the front slave nodes.

Have any one had any similar problems?