A TRANSPARENT MAC METHOD FOR BANDWIDTH SHARING AND CDV CONTROL AT THEATM LAYER OF PASSIVE OPTICAL NETWORKS

Citation
Jd. Angelopoulos et al., A TRANSPARENT MAC METHOD FOR BANDWIDTH SHARING AND CDV CONTROL AT THEATM LAYER OF PASSIVE OPTICAL NETWORKS, Journal of lightwave technology, 14(12), 1996, pp. 2625-2635
Citations number
20
Categorie Soggetti
Optics
ISSN journal
07338724
Volume
14
Issue
12
Year of publication
1996
Pages
2625 - 2635
Database
ISI
SICI code
0733-8724(1996)14:12<2625:ATMMFB>2.0.ZU;2-Y
Abstract
In the quest fora way to bring local loop costs down and make B-ISDN a ffordable to small traffic customers, the passive optical networks (PO N's) in combination with ATM (APON's) have been investigated extensive ly, To achieve the required medium sharing on a time division multiple xing (TDMA) basis, an arbitration mechanism is required in the upstrea m direction, The resulting multiplexing and concentration, though it i s the source of significant savings, it has been plagued by the penalt y of high distortion of the traffic profiles at the exit of the PON, T he preventive congestion control method of ATM is rather intolerant of such distortions in the multiplexing stages, Arbitration methods whic h are either semistatic or based on number of arrivals per reservation period, cause very high cell delay variation (CDV). The cell clusteri ng introduced by the arbitration method cannot be distinguished from p eak rate violations of the contract by the policing unit, The result i s either unwarranted rejections of conforming cells or unacceptably lo w utilization, The scheme presented herewith maintains a CDV behavior comparable to that of a common centralized FIFO multiplexer, It relies on a meticulous recording of the arrival timing of cells enabling a s ubsequent allocation of slots which produces almost the same output (a nd CDV) as the centralized FIFO multiplexer. A compact coding of the t iming information restricts bandwidth waste for control information to the size found in other similar protocols, The most important feature is that the scheme is amenable to easy hardware implementation and do es not require any call related information.