Optimal and near-optimal alpha-fair resource allocation algorithms based on traffic demand predictions for optical network planning

Abstract

We examine proactive network optimization in reconfigurable optical networks based on traffic predictions. Specifically, the fair spectrum allocation (SA) problem is examined for a priori reserving resources aiming to achieve near-even minimum quality-of-service (QoS) guarantees for all contending connections. The fairness problem arises when greedy SA policies are followed, which are based on point-based maximum demand predictions, especially under congested networks, with some connections highly overprovisioned and others entirely blocked, resulting in highly uneven QoS connection guarantees. To address this problem, we consider predictive traffic distributions allowing the exploration of several combinations of possible SAs. To find a fair SA policy, we resort to an α-fairness scheme, while QoS fairness is evaluated according to a game-theoretic analysis based on the coefficient of variations of the connections' unserved traffic metric, which measures the dispersion of unserved traffic for a connection around the mean of the unserved traffic over all connections. Non-contending (NC) and link-based contending (LBC) α-fair SA integer linear programming algorithms are proposed, where the optimal NC approach encompasses global network contention information, while the near-optimal LBC approach encompasses partial contention information to reduce problem complexity. We show that as parameter α increases, QoS fairness improves, along with connection blocking, resource utilization, and overprovisioning and underprovisioning. Further, LBC exhibits results close to the optimal, with a significant improvement in processing time.

A Reward-Based Fair Resource Allocation in EONs Considering Traffic Demand Behavior

“…In this work, two reward functions are examined. The first reward function r(u, max f ) depends on the amount of allocated utilities u and the peak-rate demands max f ∈ R n of the connections, leading to the optimization objective of [5]. The second reward function, r(u, f, θ), depends on the amount of allocated utilities u, fluctuations f = [f 1 , ..., f n ] drawn by traffic demand distributions F = [F 1 (•), .., F n (•)] as f ∼ F , and the parameter θ that defines a tolerance amount of unserved traffic for all connections.…”

“…By doing so, SAs that lead to connection blocking (u ij = 0) or unnecessary over-provisioning (i.e., u ij > max f i ) are less preferred compared to the SAs with scores that increase as u ij increases. Furthermore, this reward function normalizes the utilities, leading to the optimization objective examined in [5]. Normalization of the utilities captures the fact that the amount of utility achieved by a connection is less important than the fraction of the maximum possible utility that the connection achieves [25]; an important consideration, especially in the network environment considered, where the approximation of an egalitarian allocation may be preferable by the network operator, to ensure that contending connections with lower bandwidth demand tendencies are allocated the highest possible SAs.…”

“…As also discussed in [5], for reducing the problem's complexity, the connections' set Λ = {1, , ..., n} is decomposed into a number of non contending sets…”

“…Recently our work in [4], [5] also investigated fair resource allocation in optical networks, utilizing the α-fairness technique [6] (with its associated social elasticity welfare function parameterized by a scalar α ≥ 0 (the inequality aversion parameter)). This work extends the aforementioned works as it utilizes a reward-based objective that it is more general than the one presented in [5].…”

“…Recently our work in [4], [5] also investigated fair resource allocation in optical networks, utilizing the α-fairness technique [6] (with its associated social elasticity welfare function parameterized by a scalar α ≥ 0 (the inequality aversion parameter)). This work extends the aforementioned works as it utilizes a reward-based objective that it is more general than the one presented in [5]. Specifically, it allows the consideration of several, differently defined, reward functions by formulating the RSA problem based on the maximization of the constant elasticity social welfare function W α (r), where r(u, F (•)) ∈ R n + is the reward received by the connections when allocated u spectrum slots (utilities), given their a-priori modeled traffic distributions F (•) ∈ R n .…”