doi:10.1016/j.peva.2003.10.008
Copyright © 2003 Elsevier B.V. All rights reserved.
Performance analysis of threshold-based selective drop mechanism for high performance packet switches
References and further reading may be available for this article. To view references and further reading you must
purchase this article.
Jui-Pin Yang
, 
Computer & Communications Research Labs, Industrial Technology Research Institute, 15-7 Fengshi Street, Shiaugang District, Kaohsiung, Taiwan, ROC
Received 29 January 2003;
Revised 2 October 2003.
Available online 24 January 2004.
Abstract
In this paper, a simple and efficient threshold-based selective drop (TSD) buffer management scheme is proposed to improve the overall loss performance and fairness by regulating the buffer sharing in a packet switch. The key idea of TSD is to divide the output ports into active or inactive output ports by comparing their queue lengths with a fair share buffer size. When total queue length equals or exceeds the threshold value, only arrival packets destined to inactive output ports are admitted to enter the shared buffer. A transient analysis of TSD is derived to prove the fairness of buffer allocation among different output queues. We use computer simulation to compare the overall loss performance of TSD, complete sharing (CS), complete partitioning (CP), dynamic threshold (DT) and pushout (PO). The TSD method is shown to be more robust than CS, CP and DT methods under different traffic conditions.
Author Keywords: Threshold; Selective drop; Buffer management; Fairness; Robust
Fig. 1. Overall loss probability vs. TSD factor under different number of heavily loaded output ports.
Fig. 2. Overall loss probability vs. number of heavily loaded output ports, for various schemes.
Fig. 3. Overall loss probability vs. various loads offered to a heavily loaded output port, for various schemes.
Fig. 4. (a) Overall packet loss probability vs. TSD factor for various number of heavily loaded output ports; (b) overall packet loss probability vs. DT factor for various number of heavily loaded output ports.
Fig. 5. (a) Overall packet loss probability vs. TSD factor for various number of heavily loaded output ports; (b) overall packet loss probability vs. DT factor for various number of heavily loaded output ports.
Fig. 6. Overall packet loss probability vs. TSD factor for various burst lengths of “ON–OFF” input traffic model.
Fig. 7. (a) Overall packet loss probability vs. TSD factor at 8 ports; (b) overall packet loss probability vs. TSD factor at 32 ports; (c) overall packet loss probability vs. TSD factor at 64 ports.
Corresponding author. Present address: 160 San-Hsing, Ming-Hsiung, , Chia-Yi 621, , Taiwan, ROC. Tel.: +886-7-8714107.