Performance Analysis of a Dynamic Bandwidth Allocation Algorithm in a Circuit-Switched Communications Network.

摘要

Military communications networks typically employ a gateway multiplexer to aggregate all communications traffic onto a single link. These multiplexers typically use a static bandwidth allocation method via time- division multiplexing (TDM). Inefficiencies occur when a high-bandwidth circuit, e.g., a video teleconferencing circuit, is relatively inactive rendering a considerable portion of the aggregate bandwidth wasted while inactive. Dynamic bandwidth allocation (DBA) reclaims unused bandwidth from circuits with low utilization and reallocates it to circuits with higher utilization without adversely affecting queuing delay. The proposed DBA algorithm developed here measures instantaneous utilization by counting frames arriving during the transmission time of a single frame on the aggregate link. The maximum calculated utilization observed over a monitoring period is then used to calculate the bandwidth available for reallocation. A key advantage of the proposed approach is that it can be applied now and to existing systems supporting heterogeneous permanent virtual circuits. With the inclusion of DBA, military communications networks can bring information to the warfighter more efficiently and in a shorter time even for small bandwidths allocated to deployed sites. The algorithm is general enough to be applied to multiple TDM platforms and robust enough to function at any line speed, making it a viable option for high-speed multiplexers. The proposed DBA algorithm provides a powerful performance boost by optimizing available resources of the communications network. Utilization results indicate the proposed DBA algorithm significantly out-performs the static allocation model in all cases. The best configuration uses a 65536 bps allocation granularity and a 10 second monitoring period. Utilization gains observed with this configuration were almost 17% over the static allocation method. Queuing delays increased by 50% but remained acceptable, even for realtime traf7.