On the Early Release of Burst-Control Packets in Optical Burst-Switched Networks

Abstract

Esta es la versión de autor de la comunicación de congreso publicada en: This is an author produced version of a paper published in: Abstract. In Optical Burst-Switched networks, the so-called BurstControl Packet is sent a given offset-time ahead of the optical data burst to advertise the imminent burst arrival, and reserve a time-slot at each intermediate node to allocate the incoming optical burst. This work proposes a methodology to estimate the number of packets to arrive in a given amount of time, in order to make it possible to send the BCP packet straightafter the first packet arrival and reduce the latency experienced during the burst-assembly process. The following studies the impact of a wrong guess in terms of overreservation of resources and waiting-time at the assembler, providing a detailed characterisation of their probability density functions. Additionally, a case example in a scenario with non-homogeneous Poisson arrivals is analysed and it is shown how to choose the appropriate burst-assembly algorithm values to never exceed a given over-reservation amount.

Analysis of average burst-assembly delay and applications in proportional service differentiation

“…In this light, since the majority of delay is due to the burst-assembly process and the offset time, and the latter is typically set by the network topology and cannot be modified, the only possible way to address delay-based class differentiation is to make use of the burst assembly process. Furthermore, the offset delay can be totally or partially removed just by sending the BCP packet with an estimate of the final burst size and expected release time, even before the data burst is actually completed (see [21,22]). To the best of the authors' knowledge, only [10] proposes a mechanism to bound the maximum burst assembly delay experienced by packets that belong to different traffic classes.…”