Dynamic Constraint-based Influence Framework and its Application in Stochastic Modeling of Load Balancing

TL;DR

This paper introduces the Dynamic and Constraint-based Influence Model (DCIM), extending existing network influence models to better capture dynamic and constrained interactions, and applies it to optimize load balancing in computing networks.

Contribution

The paper proposes the DCIM, a novel extension of the Influence Model, enabling modeling of dynamic, constraint-based network interactions and applying it to load balancing.

Findings

DCIM accurately predicts load distribution in networks.

DCIM identifies optimal workload distribution policies.

The model overcomes limitations of previous influence frameworks.

Abstract

Components connected over a network influence each other and interact in various ways. Examples of such systems are networks of computing nodes, which the nodes interact by exchanging workload, for instance, for load balancing purposes. In this paper, we first study the Influence Model, a networked Markov chain framework, for modeling network interactions and discuss two key limitations of this model, which cause it to fall short in modeling constraint-based and dynamic interactions in networks. Next, we propose the Dynamic and Constraint-based Influence Model (DCIM) to alleviate the limitations. The DCIM extends the application of the Influence Model to more general network interaction scenarios. In this paper, the proposed DCIM is successfully applied to stochastic modeling of load balancing in networks of computing nodes allowing for prediction of the load distribution in the system, which is a novel application for the Influence Model. The DCIM is further used to identify the optimum workload distribution policy for load balancing in networked computing systems.