Self-managed distributed mediation networks

A method for mediating the flow of network traffic in a computer network, wherein the computer network has a mediation network that comprises:

• a plurality of types of network modules forming a network topology, including:
• local points of presence (LPP) modules for receiving and transmitting network traffic between the mediation network and client programs;
• mediator (M) modules for hosting mediation tasks;
• mediator router (MR) modules for analyzing the content of incoming messages, each MR module routing the incoming messages to a predetermined mediation task in dependence upon said content;
• and,
• transmission proxy (TP) modules for forwarding messages to at least one of said LPP modules, wherein each of the MR, M and TP modules are adapted such that all paths for network traffic therethrough are non-reciprocal;
• and
• wherein a processor executes an autonomic control plane, said autonomic control plane being separate and distinct from said modules, the autonomic control plane controlling said modules
• wherein, in the method, incoming messages are propagated along a mediation cycle that comprises the steps of:
• an LPP module addressing incoming messages to a respective one of said at least one mediator router (MR) modules;
• at said addressed MR module, analyzing the content of incoming messages and routing said messages to a predetermined mediator module in dependence upon said analyzed content;
• at said predetermined mediator module, applying the mediation task to said analyzed messages and directing said mediated messages to a respective one of said TP modules;
• and
• at said TP module that receives said mediated messages, forwarding said mediated messages to at least one of said LPP modules,
• wherein the autonomic control plane effects the distribution of mediation tasks amongst the M modules, where the distribution of mediation tasks comprises the transfer of a particular mediation task from a first M module to a second M module, the transfer comprising:
• the autonomic control plane calling a HANDOVER_SEGMENT effector method on the first M module;
• the first M module, on receipt of the HANDOVER_SEGMENT effector method, changing state to a HANDOVER_SENDER state, the first M module then processing content currently stored at the first M module relating to the particular mediation task and subsequently sending a MEDIATION_CHANGE control signal to all LPP modules;
• sending a MEDIATION_CHANGE control signal from the first M module to the second M module, and forwarding content relating to the particular mediation task subsequently received by the first M module to the second M module;
• the second M module, on receipt of the MEDIATION_CHANGE control signal, changing state to a HANDOVER_RECEIVER state and sending a sensor signal to the autonomic control plane indicating this change of state, the second M module, then sending a NEW_MEDIATOR control signal to all LPP modules, wherein the LPP modules buffer content related to the particular mediation task that is received after the NEW_MEDIATOR control signal but prior to the MEDIATION CHANGE control signal;
• the autonomic control plane, on receipt of the sensor signal indicating that the second M module is in a HANDOVER_RECEIVER state, calling an effector method on the MR modules to instruct MR modules to forward content relating to the particular mediation task to the second M module rather than the first M module;
• each MR module, on changing the destination of forwarded content to the second M module, emitting a sensor signal to the autonomic control plane indicating the change and forwarding a RT_CHANGED control message for that MR module to the first M module, the first M module subsequently forwarding the RT_CHANGED control message to the second M module, wherein the second M module buffers content relating to the particular mediation task received directly from each MR module until it receives the RT_CHANGED control message for that MR module;
• the second M module emitting a sensor signal to the autonomic control plane when it receives each RT_CHANGED signal originating at each MR module;
• and
• the autonomic control plane calling effectors methods on the first and second M modules to return them to a stable state once it has received sensor signals from the second M module indicating that RT_CHANGED control signals have been received from all MR modules.