Rafael Augusto Scaraficci

A GRASP strategy for a more constrained School Timetabling Problem

This work treats a more constrained School Timetabling Problem (STP). It consists of scheduling a set of lectures and teachers in a prefixed period of time, satisfying a set of operational requirements. We applied a Greedy Randomized Adaptive Search Procedure (GRASP) heuristic, combined with a path-relinking improvement. During execution of the GRASP heuristic, the local search procedure interleaves two types of movements. In addition, the path-relinking strategy was itself enhanced by a local search procedure. The algorithms were tested with real instances and proved to be good approaches to solve this problem. Although some restrictions are specific to Brazilian educational institutions, the same ideas can inspire similar approaches for solving the STP in other situations.

Towards software defined autonomic terabit optical networks

This paper presents an Optical SDN architecture and implementation enabled for virtual optical networks, supporting adaptive and cognitive algorithms to enhance QoT. The paper also shows experimental results of a software-defined autonomic flexible transponder.

Alternative routing and zone-based spectrum assignment algorithm for flexgrid optical networks

In optical flexgrid networks, the optical spectrum can be allocated at a much finer granularity than it can be in fixed-grid WDM networks and variable spectrum widths can be assigned to the lightpaths according to the transmission rates, leading to more efficient spectrum usage. However, the dynamic establishment and tear down of lightpaths yields to the fragmentation of the spectrum with increase in blocking of connections, mainly connections demanding large spectrum widths. This paper proposes a novel algorithm that introduces a zone-based assignment policy together with an alternative path routing mechanism based on the maximum capacity available. Results derived via simulation using different topologies and considering connections requests from 40Gbps to 1000Gbps, show that the proposed algorithm leads to blocking ratio consistently lower than those given by traditional approaches. It also reduces the blocking ratio of high rates connections, in some cases, by a factor of six.

Fragmentation Aware Routing and Spectrum Assignment Algorithm

In flex-grid (elastic) networks, the spectrum can be allocated at a much finer granularity than it can be allocated in WDM networks. However, the dynamic establishment and tear down of lightpaths yields to the fragmentation of the spectrum with consequent increase in blocking of requests for connection establishment. Therefore, it is of paramount importance that allocation decisions try to mitigate the fragmentation problem. In line with that, this paper introduces the Multigraph Shortest Path Algorithm for novel Routing and Spectrum Allocation (RSA) in elastic networks. Results indicate that the joint use of the new algorithm with proposed cost functions can produce bandwidth blocking ratio four orders of magnitude lower than existing RSA algorithms.

Algorithm for energy efficient lightpath establishment in WDM networks

The problem of lightpath provisioning in WDM optical networks has been extensively studied to reduce cost and blocking, but there is still a need of investigating this problem from a power consumption perspective. In this paper, we introduce a novel algorithm for IP/MPLS over WDM networks based on a reduced auxiliary graph that limits the solution space and captures the power consumption of lightpath provisioning. Results derived via simulation demonstrate that our strategy is energy efficient and that it is able to produce acceptable blocking ratio which are close to values produced by algorithms oriented to the reduction of blocking.

Algorithm for energy efficient routing, modulation and Spectrum Assignment

Information and Communication Technology activities consumed 4% of the world energy in 2009, and such consumption will continue to increase due to the traffic growth of the Internet predicted for the next years. Techniques to make the core of the network more energy efficient has been proposed, among them, green routing has been considered a promising technique. This paper proposes a novel Routing, Modulation Level and Spectrum Assignment (RMLSA) algorithm for elastic optical networks that considers the energy consumption of potential routes. Results indicate that this algorithm can save up to 34% energy and produce bandwidth blocking ratio two orders of magnitude lower than existing energy aware RMLSA algorithms.

Traffic grooming of batches of deadline-driven requests in elastic optical networks

This paper introduces a novel traffic grooming algorithm for the connection establishment of deadline-driven requests in elastic optical networks, named Elastic Batch Grooming Algorithm. The algorithm grooms batches of requests to establish lightpaths with diverse bandwidth demands with deadline requirements. Results show that the algorithm significantly reduces the blocking ratio and the number of demanded transponder when compared to traditional non-batch algorithms.

Hybrid Heuristic Strategies for Planning and Scheduling Forest Harvest and Transportation Activities

This work treats a forest planning problem stemming from the daily operation of some large pulp and paper companies. The problem consists in planning and scheduling harvest and transportation activities for each day in a planning horizon of around one year, while satisfying a sizable and complex set of constraints related to the structure of the harvest areas, structure and productivity of the harvest teams, transportation conditions during the rainy seasons, and some properties of the harvested logs. We propose a hybrid approach focused on GRASP (greedy randomized adaptive search procedure) metaheuristic and on linear models. Other techniques such as construction based on memory, path-relinking, and recombination methods were used to enhance the basic GRASP procedure. These strategies were tested on real field data and also with some newly generated instances. They proved to be an adequate approach in all cases.