Modifications to and performance evaluation of output scheduling in 3LIHON nodes
Optical networks became very important in the latest years, thanks to
their high trac capacity, because of the continuous growth of total
exchanged data and the request of real-time communication.
Since they have (relatively) long adaptation times, this technology is
not inherently the most suitable to carry bursty traffic like the current
one. Hybrid networks try to get the best from optical circuit switching
and optical packet switching.
This work is focused on an hybrid network architecture called 3LIHON (3-Level Integrated Hybrid Optical Network). It has three different quality of service (QoS) levels, in order to meet different requirements:
- Guaranteed Service Type (GST): resembling an optical circuit
switched service, it does not allow data loss.
- Statistically Multiplexed Real Time (SM/RT): resembling an
optical packet switched service, it ensures none or very small
delay inside the network, it allows a low data loss ratio and
- Statistically Multiplexed Best Effort (SM/BE): resembling an
optical packet switched service with very small overall packet
loss but no guaranteed delay inside the nodes.
In a 3LIHON node, SM/BE traffic interrupted by packets with higher
priority is dropped. This means that time and resources spent sending
the SM/BE packet until the interruption are wasted.
In this work we try to avoid this behaviour, by implementing and
comparing three new output scheduling algorithms. They are different
versions of the standard 3LIHON node:
- 3LIHON-RS: SM/BE interrupted packets are resumed once
higher priority trac has been sent.
- 3LIHON-RT: SM/BE interrupted packets are retransmitted.
- 3LIHON-2R: since GST packets are usually much longer than
SM/RT ones, resuming an interruption caused by GST traffic
leads to a long wait. This architecture resumes SM/BE packets
interrupted by SM/RT traffic and retransmits them if they have
been interrupted by GST traffic.
A proper (Fragment End) Optical Code can be used to implement and
signal packets interruption and resuming, Every 3LIHON version has
been tested with increasing traffic loads, but never so high to overload
the node. Three different simulators have been written in order to
study the behaviour of the different architectures, by using the Simula programming language and its context class DEMOS, specifically
intended for discrete event modelling. The node input part has just
been modelled, since it does not directly influences results.
As a function of system load, performances of these new architectures
have been analysed, paying attention especially to SM/BE packets
mean delay, SM/BE queues, delay distribution of SM/BE traffic,
wavelengths utilization, GST and SM/RT interruptions over SM/BE
traffic, standard deviation of SM/BE delay and SM/RT packet loss.
It has been observed that 3LIHON-2R does not have the best absolute
performances, but is the best trade off among all the systems
and is very well balanced when system load increases. It is the architecture
showing the lowest mean SM/BE packet delay. 3LIHON-RS proved to have the lowest SM/BE queue values and less interruptions of SM/BE traffic. Furthermore, it has the best wavelengths utilization performances. 3LIHON-RT achieved the worst results and is the first architecture overloading.
Place, publisher, year, edition, pages
Institutt for telematikk , 2013. , 126 p.
IdentifiersURN: urn:nbn:no:ntnu:diva-24243Local ID: ntnudaim:10489OAI: oai:DiVA.org:ntnu-24243DiVA: diva2:703464
Stol, Norvald, FørsteamanuensisCerroni, Walter