LINEAR LIGHTWAVE NETWORKS - PERFORMANCE ISSUES

Citation
Te. Stern et al., LINEAR LIGHTWAVE NETWORKS - PERFORMANCE ISSUES, Journal of lightwave technology, 11(5-6), 1993, pp. 937-950
Citations number
16
Categorie Soggetti
Optics
ISSN journal
07338724
Volume
11
Issue
5-6
Year of publication
1993
Pages
937 - 950
Database
ISI
SICI code
0733-8724(1993)11:5-6<937:LLN-PI>2.0.ZU;2-L
Abstract
An architecture for lightwave networks is presented based on establish ing controllable, optically transparent paths among network users. The objective is to provide high bandwidth (GigaHertz) ''clear channel'' optical connections on demand to large user populations (tens of thous ands) spread over large geographical areas (thousands of kilometers). The networks in question perform only linear operations on the optical signals (including optical amplification but excluding regeneration); hence the name ''linear lightwave network'' (LLN). Because they are c ontrollable, and are based on arbitrary topologies they are capable of being reconfigured dynamically in response to changing load condition s or component failures. The LLN's operate in a wavelength division mu ltiplexed mode. Each network node consists of a linear divider combine r (LDC), which operates as a generalized optical switch, performing co ntrollable, waveband selective optical signal routing, combining and s plitting. The network stations consist of tunable transmitters and rec eivers, and are connected to a network node through a pair of access f ibers. Connections are established by selecting an optical transmissio n channel, creating an optical path for the channel by the switching a ction of the LDC's, and tuning the participating transmitter and recei ver(s) to the assigned channel. We focus on performance issues in LLN' s subject to random demand, showing through illustrative examples how the traffic handling capability of a network is influenced by its topo logy, by the constraints imposed by technological limitations, and by the effectiveness of the routing and channel assignment algorithms. St atic and dynamic routing techniques are considered and their performan ce is compared. The static routing methods utilize preassigned wavelen gth routing to achieve a high degree of reuse of the optical spectrum in different parts of the network. In the dynamic routing algorithms t his reuse is achieved as a natural result of the dynamic behavior of t he algorithm. Quantitative results are obtained for blocking probabili ty as a function of offered load and as a function of technological co nstraints (degree of wavelength selectivity in the LDC's and number of available channels). These results are obtained analytically for the static routing case and via simulation for the dynamic routing case. I t is shown that without exceeding the limits of currently available te chnology, LLN's with the order of one thousand nodes and ten thousand stations, capable of throughputs in the range of 10s of Terabits per s econd are feasible.