Advanced Computation Capacity Modeling for Delay-Constrained Placement of IoT Services

Abstract

A vast range of sensors gather data about our environment, industries and homes. The great profit hidden in this data can only be exploited if it is integrated with relevant services for analysis and usage. A core concept of the Internet of Things targets this business opportunity through various applications. The virtualized and software-controlled 5G networks are expected to achieve the scale and dynamicity of communication networks required by Internet of Things (IoT). As the computation and communication infrastructure rapidly evolves, the corresponding substrate models of service placement algorithms lag behind, failing to appropriately describe resource abstraction and dynamic features. Our paper provides an extension to existing IoT service placement algorithms to enable them to keep up with the latest infrastructure evolution, while maintaining their existing attributes, such as end-to-end delay constraints and the cost minimization objective. We complement our recent work on 5G service placement algorithms by theoretical foundation for resource abstraction, elasticity and delay constraint. We propose efficient solutions for the problems of aggregating computation resource capacities and behavior prediction of dynamic Kubernetes infrastructure in a delay-constrained service embedding framework. Our results are supported by mathematical theorems whose proofs are presented in detail.