Banca de DEFESA: MATHEUS LÔBO DOS SANTOS

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : MATHEUS LÔBO DOS SANTOS
DATE: 10/12/2021
TIME: 09:00
LOCAL: Sessão por Videoconferência - Sala: meet.google.com/wpf-earg-rpf
TITLE:

Approaches to Spectrum Assignment in Elastic Optical Networks Based on Capacity Loss Under Multiple Routes.

KEY WORDS:

Elastic Optical Networks. Spectrum Allocation. MSCL.

PAGES: 69
BIG AREA: Engenharias
AREA: Engenharia Elétrica
SUMMARY:

This Dissertation discusses solutions to the problem of routing and spectrum allocation in elastic optical networks. The presented proposals are an adaptation of the spectrum allocation algorithm known as Min Slot Continuty Loss (MSCL). This algorithm reduces the impact on the network capacity by choosing a set of slots that provides the minimum loss of capacity. Both proposals use a set of multiple routes obtained by the Yen-MH routing algorithm, which routes are selected based on the smallest number of links. The heuristics are called MSCL Sequencial and MSCL Combinado, relating how they choose the solution among those with the lowest loss of capacity on the routes. MSCL Sequencial and MSCL Combinado were compared with the widely used heuristic in the literature referred to as First-Fit, which requires low computational effort and presents good results. The results show the capacity of the MSCL heuristics to reduce the blocking probability, presenting the following order in performance: First-Fit, MSCL Sequencial and MSCL Combinado. However, the MSCL algorithm demands high computational effort, therefore, optimization procedures were applied to improve the simulation time of the heuristics, keeping the commitment with the blocking probability. Optimizations use static heuristics (i.e., those with pre-configured ordering) and dynamic ones (i.e., those with ordering modified during network operation). A genetic algorithm based on NSGAII was developed to find more generic and well-performing solutions in the space of blocking probability and simulation time.  The heuristics were compared with the standard MSCL and improvements on blocking probability and simulation time were found. The results show the benefits of the optimization methods, demonstrating the effectiveness of their use.

BANKING MEMBERS:
Externo à Instituição - ALEX FERREIRA DOS SANTOS - UFRB
Externo à Instituição - ANDRÉ VICTOR SILVA XAVIER - UFPE
Externo à Instituição - DANILO RICARDO BARBOSA DE ARAÚJO - UFRPE
Presidente - 1963122 - RAUL CAMELO DE ANDRADE ALMEIDA JUNIOR