TY - JOUR
T1 - A parallel multiobjective approach based on honey bees for traffic grooming in optical networks
AU - Rubio-Largo, Álvaro
AU - Vega-Rodríguez, Miguel A.
AU - González-Álvarez, David L.
N1 - Rubio-Largo, Á., Vega-Rodríguez, M. A., & González-Álvarez, D. L. (2015). A parallel multiobjective approach based on honey bees for traffic grooming in optical networks. Computer Journal, 58(9), 2171-2191. https://doi.org/10.1093/comjnl/bxv046
PY - 2015/9/29
Y1 - 2015/9/29
N2 - In this paper, we propose a parallel multiobjective approach based on the honey bees' behaviour for grooming low-speed traffic requests onto high-capacity optical channels. This problem of grooming low-speed traffic requests is known in the literature as the traffic grooming problem. This problem in mesh optical networks is an NP-hard problem, in which the solution time raises exponentially when the network size increases; therefore, the use of metaheuristics and parallelism jointly is a great option in order to reduce the prohibitive runtime. The parallel approach is implemented for shared-memory architectures by using OpenMP. In our experiments, we have measured the speedup and efficiency of the method with 1, 2, 4, 8 and 16 cores when dealing with three optical networks: small (6-node network, 6 nodes), medium (National Science Foundation, 14 nodes) and large (Nippon Telegraph and Telephone, 55 nodes). Furthermore, we present a comparative study with traditional methods and other multiobjective metaheuristics published in the literature; in which we show the advantages of our multiobjective proposal.
AB - In this paper, we propose a parallel multiobjective approach based on the honey bees' behaviour for grooming low-speed traffic requests onto high-capacity optical channels. This problem of grooming low-speed traffic requests is known in the literature as the traffic grooming problem. This problem in mesh optical networks is an NP-hard problem, in which the solution time raises exponentially when the network size increases; therefore, the use of metaheuristics and parallelism jointly is a great option in order to reduce the prohibitive runtime. The parallel approach is implemented for shared-memory architectures by using OpenMP. In our experiments, we have measured the speedup and efficiency of the method with 1, 2, 4, 8 and 16 cores when dealing with three optical networks: small (6-node network, 6 nodes), medium (National Science Foundation, 14 nodes) and large (Nippon Telegraph and Telephone, 55 nodes). Furthermore, we present a comparative study with traditional methods and other multiobjective metaheuristics published in the literature; in which we show the advantages of our multiobjective proposal.
KW - honey bees
KW - multiobjective optimization
KW - OpenMP
KW - traffic grooming
UR - http://www.scopus.com/inward/record.url?scp=84940702520&partnerID=8YFLogxK
U2 - 10.1093/comjnl/bxv046
DO - 10.1093/comjnl/bxv046
M3 - Article
AN - SCOPUS:84940702520
VL - 58
SP - 2171
EP - 2191
JO - Computer Journal
JF - Computer Journal
SN - 0010-4620
IS - 9
ER -