Comparative Performans Analysis of Slotted Optical Burst and Packet Switching Techniques
|
|
Optical networks allow the solution of many problems seen in current computer networks. In addition to providing a very high capacity, they also provide a common network infrastructure that supports various services. Also in optical networks, bandwidth can be adjusted in a flexible manner according to need. In the case of the bandwidth requirements, basically three solutions arise. These are OCS (Optical Circuit Switching), OPS (Optical Packet Switching) and OBS (Optical Burst Switching). Optical switches and networks can be divided into two categories that are slotted and unslotted. Slotted is based on fixed-length time slots and synchronous packet processing. In unslotted, the packet length is variable. In SOBS (Slotted Optical Burst Switching), burst length is fixed. Routers are synchronized and they only send bursts at the beginning of the time slot. In SOPS networks, time is slotted and switch structure in each node can be configured again at the beginning of a time slot. All packets in this network are the same size and length of a slot is equal to sum of optical packet size and length of optical header.
In this thesis study, firstly the studies that have been made so far about SOBS and SOPS techniques are examined. After that SOBS and SOPS methods used for comparison study are explained. Theoretical analyses of these methods are given and algorithmic analyses are prepared. In this study, the comparative performance analysis of SOBS and SOPS techniques are performed with simulation studies. Simulations are made for different number of links and wavelengths on NSFNET and ring topologies. Drop probabilities, access delays and end to end delays are considered as performance criteria.
According to the simulation results, it is observed that SOBS gives better results than SOPS in terms of loss probability in both topologies. The loss probabilities of SOBS and SOPS are better in NSFNET topology. As seen in the simulation results, the increase in traffic load leads to the reduction of access delays on NSFNET and ring topologies for SOBS. In SOPS, when the traffic load increases, access delays also increase. It is seen that SOPS and SOBS give better results in NSFNET topology than ring topology in terms of access delay. In low traffic load, SOPS has lower access delays than SOBS for both topologies. When the traffic load increases, access delay of SOPS on ring topology is greater than access delay of SOBS on both topologies. In addition to that access delay of SOPS on NSFNET topology is the lowest in all traffic loads. In other words, access delays of SOPS are more in ring topology. According to the simulation results, it is seen that end to end delays of SOBS and SOPS are less in ring topology. End to end delay of SOPS is less in NSFNET topology. In ring topology, while the traffic load is low, end to end delay of SOPS is less. In contrast, end to end delay of SOBS is less while the traffic load is high.
In this thesis, it is shown that choosing one of these two techniques depends on the traffic load of network and the topologies. Finally, some information is given about possible subjects to study in the later stages of this work.
|
BATUR Özlem
Danışman : Prof. Dr. Ahmet SERTBAŞ
Anabilim Dalı : Bilgisayar Mühendisliği
Programı : -
Mezuniyet Yılı : 2014
Tez Savunma Jürisi : Prof. Dr. Ahmet SERTBAŞ
Doç. Dr. Fırat KAÇAR
Doç. Dr. Hakan DOĞAN
Yard. Doç. Dr. Tolga ENSARİ
Yard. Doç. Dr. Fatih KELEŞ
Dostları ilə paylaş: |