Research: Space division approach to implement a shared buffer in an ATM switch

Authors:
Derek C. W. Pao;S. C. Leung
Affiliations:
Department of Electronic Engineering, City University of Hong Kong, Tat Chee Avenue, Kowloon, Hong Kong;Department of Computing Studies, Hong Kong Baptist University, Waterloo Road, Kowloon, Hong Kong
Venue:
Computer Communications
Year:
1997

Citing 7
Cited 0

Analysis of input and output queueing for nonblocking ATM switches

IEEE/ACM Transactions on Networking (TON)
ATM switch architectures with input-output-buffering: effect of input traffic correlation, contention resolution policies, buffer allocation strategies and delay in backpressure signal

Computer Networks and ISDN Systems
A flexible shared-buffer switch for ATM at Gb/s rates

Computer Networks and ISDN Systems
Design of ATM switch using hypercube with distributed shared input buffers and dedicated output buffers

ICNP '95 Proceedings of the 1995 International Conference on Network Protocols
A survey of modern high-performance switching techniques

IEEE Journal on Selected Areas in Communications
Sunshine: a high performance self-routing broadband packet switch architecture

IEEE Journal on Selected Areas in Communications
Queueing in high-performance packet switching

IEEE Journal on Selected Areas in Communications

Quantified Score

Hi-index	0.24

Visualization

Abstract

In this paper, we propose a space division approach to implement a shared buffer in an ATM switch which does not require scaling up the bandwidth of the shared memory. At the normal operating point, cell losses are infrequent, and overloading of individual output ports is likely to occur at different time instances. An external 'shared' buffer can be time-multiplexed to serve the overloaded output ports on demand. To achieve this, the switch provides an additional feedback path for derouting cells destined for an overloaded output port to the shared buffer. When the traffic load at the corresponding output port drops and buffer space becomes available in the output queue, the derouted cells will be removed from the shared buffer and routed via the switching fabric to their destinations. Simulation results reveal that by incorporating a shared buffer of size 128, the acceptable load of the switch can be substantially increased, especially for bursty traffic conditions.