Benjamín Barán

Solving multiobjective multicast routing problem with a new ant colony optimization approach

This work presents two multiobjective algorithms for Multicast Traffic Engineering. The proposed algorithms are new versions of the Multi-Objective Ant Colony System (MOACS) and the Max-Min Ant System (MMAS), based on Ant Colony Optimization (ACO). Both ACO algorithms simultaneously optimize maximum link utilization and cost of a multicast routing tree, as well as average delay and maximum end-to-end delay, for the first time using an ACO approach. In this way, a set of optimal solutions, know as Pareto set is calculated in only one run of the algorithms, without a priori restrictions. Experimental results show a promising performance of both proposed algorithms for a multicast traffic engineering optimization, when compared to a recently published Multiobjective Multicast Algorithm (MMA), specially designed for Multiobjective Multicast Routing Problems.

Multi-objective optimization scheme for multicast flows: a survey, a model and a MOEA solution

This paper presents a new traffic engineering load balancing taxonomy, classifying several publications and including their objective functions, constraints and proposed heuristics. Using this classification, a novel Generalized Multiobjective Multitree model (GMM-model) is proposed. This model considers for the first time multitree-multicast load balancing with splitting in a multiobjective context, whose mathematical solution is a whole Pareto optimal set that can include several results than it has been possible to find in the publications surveyed. To solve the GMM-model, a multi-objective evolutionary algorithm (MOEA) inspired by the Strength Pareto Evolutionary Algorithm (SPEA) is proposed. Experimental results considering up to 11 different objectives are presented for the well-known NSF network, with two simultaneous data flows.

Routing and wavelength assignment over WDM optical networks: a comparison between MOACOs and classical approaches

The increasing demand of bandwidth has found an answer in Optical Transport Networks (OTN). To take advantage of the different resources that OTNs offer, several parameters need to be optimized to obtain good performance. Therefore, this work studies the Routing and Wavelength Assignment (RWA) problem in a multiobjective context. MultiObjective Ant Colony Optimization (MOACO) algorithms are implemented to simultaneously optimize the hop count and number of wavelength conversion for a set of unicast demands, considering wavelength conflicts. This way, a set of optimal solutions, known as Pareto Set, is calculated in one run of the proposed algorithm, without a priori restrictions on some objective. The proposed MOACO algorithms were compared to classical RWA heuristics using several performance metrics. Although, there is not a clear superiority, simulation results indicate that considering most of the performance metrics, MOACO algorithms obtain promising results when compared to the classical heuristics.

Generalized multiobjective multitree model for dynamic multicast groups

Generalized multiobjective multitree model (GMM-model) studied for the first time multitree-multicast load balancing with splitting in a multiobjective context. To solve the GMM-model, a multiobjective evolutionary algorithm (MOEA) inspired by the strength Pareto evolutionary algorithm (SPEA) was already proposed. In this paper, we extend the GMM-model to dynamic multicast groups (i.e. egress nodes can change during the connections lifetime), given that, if recomputed from scratch, it may consume a considerable amount of CPU time. To alleviate this drawback we propose a dynamic generalized multiobjective multitree model (dynamic-GMM-model) that in order to add new egress nodes makes use of a multicast tree previously computed with the GMM-model. To solve the dynamic-GMM-model, a new MASPA (multiobjective approximation using shortest path algorithm) heuristic is proposed. Experimental results considering up to 11 different objectives are presented for the well-known NSF network. We compare the performance of the GMM-model using MOEA with the proposed dynamic-GMM-model using MASPA, showing that reasonable good solutions may be found using fewer resources (such as memory and time). The main contributions of this paper are the optimization model for dynamic multicast routing; and the proposed heuristic algorithm.

Multitree-multiobjective multicast routing for traffic engineering

This paper presents a new traffic engineering multitree-multiobjective multicast routing algorithm (M-MMA) that solves for the first time the GMM model for Dynamic Multicast Groups. Multitree traffic engineering uses several trees to transmit a multicast demand from a source to a set of destinations in order to balance traffic load, improving network resource utilization. Experimental results obtained by simulations using eight real net-work topologies show that this new approach gets trade off solutions while simultaneously considering five objective functions. As expected, when M-MMA is compared to an equivalent singletree alternative, it accommodates more traffic demand in a high traffic saturated network.

Hashing based traffic partitioning in a multicast-multipath MPLS network model

Load Balancing is a key mechanism in traffic engineering. One interesting strategy for load balancing enhancement is the multipath approach, in which data is transmitted through different paths. The use of effective hashing functions for load balancing optimizes the network utilization and reduces packet disordering and imbalance. This paper address the problem of packet ordering in multipath - multicast MPLS networks, studies the impact of the hashing function to effectively partition the traffic to implement the flow splitting values issued from an optimized model and analyzes the traffic allocation to the LSPs of the network and the mis-ordering problem at the egress node using buffer schemes. The buffer allocation levels are calculated according to end-to-end delays. Finally, the paper presents some experimental results from an optimized network.

A Comparative Evaluation of Algorithms for Auction-Based Cloud Pricing Prediction

Nowadays, cloud computing providers offer idle resources through an auction-based system in order to maximize resource utilization and economical revenue. Cloud computing consumers have the opportunity to take advantage of the resources offered at very low spot price in exchange for lower reliability in the provision of these resources. In this context, the Spot Price Prediction (SPP) is a well studied problem mainly formulated as a time series prediction, with particularities of auction-based cloud markets. This work presents a comparative evaluation of three different well-known prediction algorithms, applied for the first time to the SPP problem, against astate-of-the-art Three-Layer Perceptron (TLP) algorithm. In order to measure the accuracy of the evaluated algorithms, the following error metrics were considered: (1) Mean-Squared Error (2) Maximum Positive Error and (3) Mean Positive Error. Experimental results indicate that the Support Vector Poly Kernel Regression (SMOReg) algorithm outperforms other evaluated algorithms for the SPP problem, improving probabilities of obtaining resources in a highly dynamic spot market.

Workload generation for virtual machine placement in cloud computing environments

Cloud computing datacenters provide millions of virtual machines (VMs) in actual cloud markets. Nowadays, efficient location of these VMs into available physical machines (PMs) represents a research challenge, considering the large number of existing formulations and optimization criteria. Several techniques have been studied for the Virtual Machine Placement (VMP) problem. However, each article performs experiments with different datasets, making difficult the comparison between different formulations and solution techniques. Considering the absence of a highly recognized and accepted benchmark to study the VMP problem, this work proposes and implements a Workload Generator to enable the generation of different instances of the VMP problem for cloud computing environments, based on different configurable parameters. Additionally, this work also provides a set of pre-generated instances of the VMP that facilitates the comparison of different solution techniques of the VMP problem for the most diverse dynamic environments identified in the state-of-the-art.

Global convexity in the bi-criteria Traveling Salesman problem

This work studies the solution space topology of the Traveling Salesman Problem or TSP, as a bi-objective optimization problem. The concepts of category and range of a solution are introduced for the first time in this analysis. These concepts relate each solution of a population to a Pareto set, presenting a more rigorous theoretical framework than previous works studying global convexity for the multi-objective TSP. The conjecture of a globally convex structure for the solution space of the bi-criteria TSP is confirmed with the results presented in this work. This may support successful applications using state of the art metaheuristics based on Ant Colony or Evolutionary Computation.

A multitree approach for multicast routing

This paper presents a new traffic engineering multitree-multiobjective multicast routing algorithm (M-MMA). Multitree traffic engineering uses several trees to transmit one multicast demand between a source and a set of destinations. The purpose of the M-MMA is to balance the traffic load and optimize the utilization of the network resources. For the accomplishment of the optimization goal, M-MMA proposes a local optimization procedure that finds solutions that improve the relative amount of information to be transmitted through each tree.The approach of the M-MMA is inspired in the ideas of the well-known Strength Pareto Evolutionary Algorithm (SPEA). It simultaneously optimizes six objective functions: maximum link utilization, total bandwidth consumption, total cost, hops count, average delay and maximum delay. Simulations on several network topologies prove an enhanced performance when compared to previously published results as the Multiobjective Multicast Algorithm (MMA).