(Item 8334) DLUR strategy for DLUS searches ------------------------------------------- Jim Fletcher made this proposal: I'd like to propose that we add an item to the DLUR architecture with regards to "locating a DLUS". I am afraid that we will have products that attempt to locate the DLUR "endlessly" and flood the network with broadcasts in doing so. I propose that we indicate in the architecture that products should use a "sliding" timer when locating the DLUS so that for example the first look is immediate, second at 1 minute, third at 4 minutes, then 8, 16, etc ... The current proposal is to update the DLUR architecture reference section 5.7.1 (external document; 7.7.1 in the AWP), "CPSVRMGR Session Recovery", concerning retry strategy when a Locate for a DLUS fails, as follows: The DLUR should never become "dead" but should be able to access a DLUS whenever one becomes available. It is recommended that all DLUR implementations incorporate the key concepts of: - alternating retry attempts between different DLUSs - increasing the time between retries later on - never giving up attempting to acquire a DLUS For example, a DLUR could attempt to acquire a DLUS until successful using the following retry procedure: 1) Immediately retry to the same DLUS 2) After some time t, retry to same DLUS again 3) After some time t, try the backup DLUS 4) Toggle between primary and backup up to m times 5) If still no success, toggle back and forth between primary and backup at t+n intervals forever.