Analysis of fibre-loop optical buffers with a void-avoiding schedule
The major growth of personalised video streaming and the paradigm shift
towards big data all add to the bandwidth requirements of Internet
users, urging network providers to provision them with more capacity.
While current optical fibre networks in theory offer capacities of over
10 Tbit/s per fibre, such transmission speeds are rarely achieved in
practice. This is due to the fact that optical packets need to traverse
multiple intermediary nodes on their path from origin to destination,
while it is hard to temporarily store the packets in case these nodes
are congested. A possible solution to this problem is to send the
packets into fibre delay loops when the transmission line of an
intermediary node is occupied. This buffering strategy is different from
classical buffering in the sense that once the transmission line becomes
available, a waiting packet first needs to traverse the remainder of its
delay loop before it is ready for possible transmission.
In this presentation, we analyse the stationary number of optical
packets being delayed in fibre loops simultaneously at an intermediate
node, under the assumption that a transmission is initiated as soon as
the transmission line and any packet is available for transmission (the
so-called void-avoiding schedule). We also study the stationary amount
of delay incurred by a packet before it is transmitted. The analysis
turns out to have surprising links with particular queueing systems, and
results turn out to be strikingly simple and familiar.
This presentation is based on joint work with Wouter Rogiest and Dieter
Fiems (Ghent University).