Load Balancing and Robust Dimensioning in Heterogeneous Networks

Summary Large-scale service systems, such as cloud networks, data centers
and skill-based business operations, are required to handle highly
heterogeneous and complex tasks, and thus increasingly involve servers
that are strongly tailored to perform specific types of jobs.
Such affinity relations and compatibility constraints are not captured
in classical paradigms for load balancing and dimensioning,
and raise fundamental methodological challenges.
In addition, uncertainty and variation in the total service demand
as well the distribution of load across various job types create a need
for adaptive learning algorithms to maintain high levels of efficiency
and performance.
The goal of this PhD project is to explore the impact of heterogeneity
and affinity relations on the achievable performance of load balancing
algorithms, design adaptive learning algorithms to deal with uncertain
and time-varying load patterns, and develop robust dimensioning
approaches to cover a wide range of load distributions in the presence
of compatibility constraints.
Supervisors Sem Borst (TU/e)
PhD Student

Diego Goldsztajn


Eindhoven University of Technology (TU/e)