TY - JOUR
T1 - OBS contention resolution performance
AU - Zalesky, A.
AU - Vu, H. L.
AU - Rosberg, Z.
AU - Wong, E. W M
AU - Zukerman, M.
PY - 2007/5
Y1 - 2007/5
N2 - In this paper, a performance comparison of resolution policies specific to optical burst switching is presented. A framework, based on a reduced-load approximation, is developed to estimate path blocking probabilities in an optical burst switched network of arbitrary topology, in which any combination of the following resolution policies is applied in a specific order: limited wavelength conversion, burst deflection and burst segmentation. Our framework is used to compare the relative performance of each resolution policy for two sets of ingress and egress router pairs defined on the NSF network. With respect to the assumptions considered, it is shown that limited wavelength conversion is more effective in reducing blocking relative to burst deflection if the maximum wavelength conversion radius is sufficiently large, otherwise burst deflection is more effective. Furthermore, limited wavelength conversion or burst deflection are more effective in reducing blocking relative to burst segmentation. Burst segmentation is justified as a stand alone resolution policy; however, using burst segmentation to complement another resolution policy offers only a marginal reduction in blocking. Both just-in-time and just-enough-time scheduling is analyzed. Simulations are implemented to corroborate the accuracy of our framework and extensions. © 2006 Elsevier Ltd. All rights reserved.
AB - In this paper, a performance comparison of resolution policies specific to optical burst switching is presented. A framework, based on a reduced-load approximation, is developed to estimate path blocking probabilities in an optical burst switched network of arbitrary topology, in which any combination of the following resolution policies is applied in a specific order: limited wavelength conversion, burst deflection and burst segmentation. Our framework is used to compare the relative performance of each resolution policy for two sets of ingress and egress router pairs defined on the NSF network. With respect to the assumptions considered, it is shown that limited wavelength conversion is more effective in reducing blocking relative to burst deflection if the maximum wavelength conversion radius is sufficiently large, otherwise burst deflection is more effective. Furthermore, limited wavelength conversion or burst deflection are more effective in reducing blocking relative to burst segmentation. Burst segmentation is justified as a stand alone resolution policy; however, using burst segmentation to complement another resolution policy offers only a marginal reduction in blocking. Both just-in-time and just-enough-time scheduling is analyzed. Simulations are implemented to corroborate the accuracy of our framework and extensions. © 2006 Elsevier Ltd. All rights reserved.
KW - Blocking probability
KW - Burst contention
KW - Burst segmentation
KW - Deflection routing
KW - Limited wavelength conversion
KW - Optical burst switching
KW - Reduced-load approximation
UR - http://www.scopus.com/inward/record.url?scp=33846216309&partnerID=8YFLogxK
UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-33846216309&origin=recordpage
U2 - 10.1016/j.peva.2006.06.002
DO - 10.1016/j.peva.2006.06.002
M3 - RGC 21 - Publication in refereed journal
SN - 0166-5316
VL - 64
SP - 357
EP - 373
JO - Performance Evaluation
JF - Performance Evaluation
IS - 4
ER -