S. Lekcharoen, C. Chaochanchaikul, and C. Jittawiriyanukoon (Thailand)
Backoff time computation, fuzzy control, congestion and backpressure
Congestion in network occurs when the demand exceeds the availability of network resources, leading to lower throughputs and longer delays. If congestion is not properly controlled, some sessions transported by the network may not meet their quality-of-service (QoS) requirements. When congestion builds up in a network, two general approaches are possible to cope with the shortage of buffer space. One approach is to drop incoming frames for which buffer is not available and to rely on the end-to-end protocols for the recovery of lost packets. However, Backoff time computation schemes, namely: pseudorandom backoff (PB) time, exponential backoff (EB) time and random backoff (RB) time are available and can be applied for in waiting time re arrangement in queue. They have proved to be inefficient in coping with the conflicting requirements, that is, low dropping frames and high conforming frames. This led us to explore alternative solutions based on artificial intelligence techniques, specially, in the field of fuzzy logic. In this paper, we propose a fuzzy backpressure that aims at detecting violations in parameter negotiation. We evaluate and compare the performance of fuzzy backpressure in backoff scheme, namely, fuzzy backpressure in pseudorandom backoff scheme (FBPP), fuzzy backpressure in random backoff scheme (FBPR) and fuzzy backpressure in exponential backoff scheme (FBPE). The performance of six backoff schemes have been investigated through fluctuations in telecommunication traffic streams (burst/silent type). Simulation results show that on VDSL frames, the fuzzy logic control scheme helps improve performance of our fuzzy control backpressure in backoff schemes much better than conventional backoff schemes once various types of burst/silence traffic are generated. Moreover, RB is found to be better for bandwidth sensitive requirement but regardless of QoS.
Important Links:
Go Back
