DDN Throttling Support

Too many DDN requests toward MME from SGW-C could lead to processing overload at MME. To reduce this load, MME dynamically requests SGW-C to reduce a certain percentage of DDN messages sent toward it for a given period time.

For DDN throttling, S-GW is required to drop a given percentage of DDNs over a given period of time. S-GW implements this functionality using a probabilistic algorithm at each session manager.

Whereas the conventional implementation of DDN throttling requires each session manager to share its list of pending DDNs for low-priority bearers with a central entity that would then calculate the net load of pending DDNs and then decide how many DDNs each session manager would have to drop. This implementation would require buffering of DDN messages at session manager. Also, due to distributed processing nature of software subsystem in chassis, it would require considerable amount of messaging between the session managers and the central entity (demuxmgr in case of Boxer) at regular intervals.

Implementing a probabilistic algorithm removes the need for buffering at session manager and also messaging with demuxmgr. Accuracy of probabilistic algorithm increase with increasing low ARP priority paging load at session manager. Even with lower paging load, accuracy would be fairly close to the throttling factor provided.

For non-release 10 compliant MME, SGW_C provides option to enable throttling through the CLI.

Threshold ARP values for low-priority bearer must be configured through S-GW Service Configuration. For example, if configured ARP value is 9, any bearer with ARP > 9 is considered low-priority bearer. DDN throttling is enabled through this configuration. If DDN throttling is enabled through SGW service configuration, each DDN message toward MME would contain the ARP IE.