Optimum Allocation of Distributed Service Workflows with Probabilistic Real-Time Guarantees

This paper addresses the problem of optimum allocation of distributed real-time workflows with probabilistic service guarantees over a set of physical resources. The discussion focuses on how such a problem may be mathematically formalized, in terms of both constraints and objective function to be optimized, which also accounts for possible business rules for regulating the deployment of the workflows. The presented formal problem constitutes a probabilistic admission control test that may be run by a provider in order to decide whether or not it is worth to admit new workflows into the system and to decide what the optimum allocation of the workflow to the available resources is. Various options are presented, which may be plugged into the formal problem description, depending on the specific needs of individual workflows. The presented problem has been implemented using GAMS and has been tested under various solvers. An illustrative numerical example and an analysis of the results of the implemented model under realistic settings are presented.