Abstract:
Internet traffic is growing at an explosive proportion due to ever-increasing innovations in technology. Consequently, the
massive growth in the number of Internet users has also led to the problem of network congestion. When congestion occurs, the quality
of network service can no longer be guaranteed. One of the widely-known active queue management (AQM) algorithms performed in
Internet router is the Random Early Detection (RED). RED itself certainly has a deficiency of applying same linear drop function for
varying network loads, resulting in poor quality of network service. This paper presents a new type of amendment to RED, named as
Triplex Region - Random Early Detection (TR-RED) Algorithm, to cope with the mentioned shortcoming. In TR-RED, three dropping
functions are implemented as an alternative to RED's one and only linear drop function. Differently from RED, the working principle
for the proposed TR-RED aims at providing different drop action (that is to say quadratic, linear and exponential) for unique levels of
congestion in the network. Moreover, using ns-3 network simulator, we conduct three simulation experiments to benchmark TR-RED with
two known enhanced RED-based AQM algorithms under three distinctive scenarios. Simulation results clearly indicated that TR-RED
is a promising algorithm which offers a surpassing and indeed remarkable performance gain in delay across all scenarios considered.
Interestingly, to upgrade from original RED to the proposed TR-RED only involves simple implementation effort. This is because only
little modification is needed in the dropping profile of RED's algorithm implementation.
