Performing Load Balancing under Constraints

2025-02-03

AI Summary

This paper addresses the constrained load balancing problem in communication networks, overcoming the limitations of conventional Join-the-Shortest-Queue (JSQ)-type policies—which ignore practical constraints such as bandwidth capacity and minimum queue utilization. For the first time, it systematically models the dual constraints—action-dependent (e.g., bandwidth allocation) and state-dependent (e.g., lower bound on queue utilization)—as a Constrained Markov Decision Process (CMDP). The proposed scheduling policy is theoretically grounded, guarantees strict satisfaction of all constraints, and integrates dynamic programming with policy optimization techniques. Extensive large-scale simulations demonstrate that the policy achieves 100% constraint compliance while significantly reducing average system occupancy. It attains a superior trade-off between end-to-end latency and resource utilization, thereby bridging theoretical rigor with engineering practicality.

Technology Category

Planning, Routing, and Scheduling: Scheduling under UncertaintyConstraint Satisfaction and Optimization: Distributed CSP/OptimizationSearch and Optimization: Mixed Discrete/Continuous Search

Application Category

Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystemsGraph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphsSystems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments

Abstract

Join-the-shortest queue (JSQ) and its variants have often been used in solving load balancing problems. JSQ minimizes the average system occupation, e.g., the customer's system time. In this paper, we extend the load balancing setting to include constraints that may be imposed due to the communication network. In particular, we cast the problem in the framework of constrained MDPs: this permit us to address both action-dependent constraints, such as, e.g, bandwidth constraints, and state-dependent constraints, such as, e.g., minimum queue utilization constraints. Unlike the state-of-the-art approaches in load balancing, our policies satisfy the constraints while delivering favorable results in terms of system occupancy. We derive policies that provably satisfy the constraints and evaluate their performance through extensive simulations.

Problem

• Extending load balancing with network constraints.
• Addressing action and state-dependent constraints.
• Deriving policies that ensure constraint satisfaction.

Innovation

• Constrained MDPs framework
• Action-dependent bandwidth constraints
• State-dependent queue utilization