In the simplest case, there is a single instance of the RMI Gateway, and at least two instances of the ENOVIA Live Collaboration Server, eMatrixServletRMI.jar.
The RMI Gateway and ENOVIA Live Collaboration Servers each run inside a Java2 activation daemon and have access to their own Live Collaboration kernel. For example, you can set up the Gateway and two (or more) servers with the RMI Gateway evenly distributing requests to the other processes. In this case, once the request is passed, the client deals directly with the secondary server, reducing the load on the Gateway. Optionally, instead of using a round-robin algorithm, one server may handle all requests for a certain amount of time or for a certain number of operations. It would then perform an implied quiesce function (that is, finish up currently running threads, complete RMI garbage collection, and then shut itself down). At the same time, the other server begins to handle new requests, providing uninterrupted, seamless request handling. The client's host URL should reference the port where the RMI Gateway is registered, the default being 1099. The default configuration of RMI Gateway distributes requests between servers in a round-robin fashion. To configure the Gateway to use the quiesce function, refer to Configuring the Gateway for Quiesce on Windows. |