The priority wavelength release protocol for dynamic wavelength allocation in WDM-TDMA PONs supporting random and quasi-random bursty traffic

Passive Optical Networks (PONs) are attractive fiber-based access systems providing broadband access solutions at a low cost. A hybrid Wavelength Division Multiplexing (WDM) - Time Division Multiple Access (TDMA) PON is a favorite PON configuration, in which the researchers have drawn much attention. We have analysed and proposed several protocols for dynamic wavelength allocation in a WDM-TDMA PON, by separately considering random or quasi-random input-traffic. By modelling the bursty in-service traffic as an ON-OFF model, we have proposed three protocols for random call arrivals. The same protocols have also been proposed for quasi-random call arrivals, resulting however in different evaluation conclusions (than in the case of random arriving calls). This fact gave rise to a fourth protocol, the Priority Wavelength Release (Pr-WR) Protocol, suitable for quasi-random input-traffic. According to this protocol, the service-classes accommodated in the PON are distinguished in two groups: the high and low priority groups. A connection between an Optical Node Unit (ONU) and the Optical Line Terminal (OLT) (the common link of the PON) can be terminated and the occupied wavelength is released, only when the common link is empty or holds low priority calls in state OFF. In this paper, we examine the Pr-WR protocol in the WDM-TDMA PON under a mixture of quasi-random and random arriving calls. The ON-OFF teletraffic model for the mixture of random and quasi-random arrivals is presented. We evaluate the behaviour of this protocol in comparison with the other three protocols in respect of: a) the Connection Failure Probability (due to the unavailability of a wavelength), and b) the Delay (low priority calls may delay in state OFF, until a new wavelength becomes available upon a new call arrival). For the Pr-WR Protocol, we also calculate the Call Blocking Probability (due to the restricted bandwidth of the wavelength) and the Burst Blocking Probability (burst of calls delays - n state OFF, due to temporal unavailability of wavelength bandwidth). We come to safe conclusion which is favourite for the Pr-Wr protocol.