The effect of packet traffic prediction limitations on DAMA schemes for future satellite packet networks

Future satellite-based packet networks (FSPN) will utilize a request/grant MAC protocol referred to as demand assigned multiple access (DAMA) for uplink allocations. DAMA is conceptually similar to bandwidth-on-demand (BoD) protocols used in terrestrial networks, such as 802.16 and DOCSIS, which flexibly share total bandwidth according to actual loads and usage policies. However, unique characteristics of FSPN result in challenges not faced by these protocols. DAMA request/grant cycles are on the order of seconds instead of tens of milliseconds. This makes the process one of prediction of future needs based on past traffic loading measurements rather than grants for specific packets. Over a decade of packet traffic study has shown the difficulty of fine-grained load prediction. If these requests are heavily relied upon for grant allocations, load prediction errors can result in unfair and non-deterministic terminal allocations that may violate terminal SLAs. This study assesses the predictability of packet traffic at FSPN request/grant timescales and compares the performance of DAMA approaches of varying request granularity, starting from coarse measures of inactivity and activity and focusing on meeting SLA commitments. The complexity of fine-grained approaches with respect to any potential gain is also considered. Rejecting fine-grained approaches to load prediction is shown to result in more deterministic DAMA performance and lower overhead while achieving efficient use of shared RF resources.