Scalable resources allocation for services based on distributed agent

Abstract

Current urban environments are populated with multiple and heterogeneous computing devices. Smart Object (SmOb) is the name given to everyday object embedded with computing capabilities in this environments. Moreover, the network form by these SmObs connected over the Internet, the Internet of Things (IoT), is expected to allow the urban environments to proactively support the users daily activities in the near future. Pervasive Computing (PerComp) is the branch of Computer Science taming these environments in support of the user`s needs. One of the tools to aid users is to assign SmObs in support of specific user`s requirements. This SmOb-user allocation problem vital to orchestrate PerCom environments around users.One main characteristic of the PerComp is that of the Volatility Principle (VP), stating that PerCom systems should be designed under the assumption that the set of participating users, hardware, and software is highly dynamic and unpredictable. The VP sets two main requirements to solutions for the Smob allocation problem: i) the allocation should be scalable in terms of the time taken to decide a scheduling of SmObs to agents, and ii) the allocation should be ecient as to serve as many users aspossible given available resources.This work proposes a new approach to solve the aforementioned SmObs allocation problem. The core element behind this approach is the utilization of microeconomics theory to solve the allocation as a negotiation between agents buying and agents supplying resources (SmOb`s functionalities). The approach was tested, in terms of scalability and effectiveness, against three other used in the related literature: a naive FIFO approach, an optimal Job Scheduling (JSP) approach, and an agent negotiation scheme for Web Services (WSN). As main findings, the proposed approach shows a better trade-off between scalability and effectiveness than the other tested approaches. It presents improvements between 28% and 85% on eciency over the FI...