Open Collections

Pricing and resource allocation in edge computing

University of British Columbia

Edge computing (EC) has been proposed to complement the cloud to meet the soaring traffic demand and accommodate diverse requirements of various services and systems in future networks. By distributing storage, computing, control, and networking functions closer to the edge, the emerging EC paradigm promises to deliver superior user experience and enable a wide range of Internet of Things applications. Despite the tremendous potential, EC is still in its infancy stage and many interesting open problems remain to be solved. This thesis aims to develop efficient algorithms for pricing, service placement, and resource allocation in EC. First, we consider the joint service placement, sizing, and workload allocation problem under demand uncertainty from the perspective of a service provider (SP). Specifically, we propose a novel two-stage adaptive robust optimization model to help the SP identify optimal locations for installing the service and the resource amount to purchase from each location. The optimal service placement and sizing solution can hedge against all possible realizations of the traffic demand within an uncertainty set. Hence, it enables the SP to balance the tradeoff between the operating cost and the service quality while taking demand uncertainty into account. Furthermore, the proposed scheme is less conservative than the static robust approach and more robust than the deterministic approach. Second, we introduce a new market-based framework for efficiently and fairly allocating the limited resources of geographically distributed heterogeneous edge nodes to competing services with diverse requirements and preferences. By properly pricing edge resources, the proposed framework generates a market equilibrium solution that not only maximizes the edge resource utilization but also allocates favorite resource bundles to the services given their budget constraints. We show that the equilibrium allocation is Pareto-optimal and satisfies desired fairness properties including sharing incentive, proportionality, and envy-freeness. We further generalize the market model to tackle the case of net profit maximization and to capture practical system design aspects such as multiple resource types and limited demand. Finally, we present distributed algorithms for equilibrium computation while respecting user privacy. Numerical results demonstrate the superior performance of the proposed market models compared to benchmark schemes.

Text

2020-08-07

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/75385

Electrical and Computer Engineering

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/