15 Proc. of the Fifth International Conference on Advances in Computing, Electronics and Communication - ACEC 2017. Copyright © Institute of Research Engineers and Doctors. All rights reserved. ISBN: 978-1-63248-121-4 doi: 10.15224/ 978-1-63248-121-4-04 Alternative Resource Allocation Model for Dynamic Resource-Sharing WSN Systems Güngör YILDIRIM Yetkin TATAR Abstract—One of the major solutions to the interoperability problem of different wireless sensor networks (WSNs) which are in different locations is the use of WSN virtualization techniques. In these types of systems, there is generally a provider system to which the heterogeneous WSNs are registered. Through the provider system, clients can choose different resources existing in the different WSNs and establish their own virtual networks. In virtualization, the resource choice/allocation operation is one of the initial processes. This process is important to clients, the provider and efficient system maintenance. Current resource choice/allocation models used in IoT WSN providers generally take into account client parameters. However, other parameters are also available for optimum resource allocation, including available energy levels of the nodes in the sub-WSNs, data traffic in the sub-WSNs, processing rate of the physical nodes. In addition, the sub-WSNs registered to the provider can have different skills and features. The sub-WSN parameters must be taken into account in the process of the resource choice/allocation. In the paper, an alternative resource allocation model which considers these critical parameters is proposed. In the model, which is analytically explained, both client-side and provider-side parameters are considered. Thus, the systems in the sub WSNs can operate for a longer time and more economically Keywords—WSN, virtualization, IoT, resource sharing I. Introduction WSN technologies have made a lot of progress for decades, and many software/hardware WSN technologies have been developed in this period. Consequently, big IoT projects such as smart city, smart grids have started using WSNs as basic infrastructure units. This has led to the emergence of immense heterogeneous structures under the IoT roof. As a result, researchers and engineers have had to find solutions to the resource sharing problems of heterogeneous WSN systems. One of the best-known solutions to the problem is WSN virtualization. Virtualization enables different applications to utilize the shared resources available in a system by hiding the infrastructure complexity. It is necessary to evaluate the WSN virtualization methods differently from other well-known virtualization methods of computer technologies. Güngör YILDIRIM Firat University, Engineering Faculty, Computer Engineering Dept. Elazig,Turkey Yetkin TATAR Firat University, Engineering Faculty, Computer Engineering Dept. Elazig,Turkey The reason for this is that WSNs, by nature, have many restrictions. Especially, critical parameters such as energy and memory limitations, low processing capacity, the large number of nodes must be taken into account at all times. Moreover, this is an optimization problem for WSN virtualization. In general, different types of WSN virtualization methods exist, node-based virtualization (NodBV), network-based virtualization (NetBV) [1-3]. But the methods are not suitable for IoT systems that meet the demands of a large number of clients. For this, middleware based sensor cloud systems are more efficient and effective. Virtualization on cloud systems completely isolates the infrastructure from the clients. Thus, the clients are only interested in resource choice [3-6]. In IoT-WSN provider systems can have different working models. For example, in a sensor cloud system, the periodic information from the nodes is transferred to the interested clients through middleware technologies. In this case, the increase in the number of clients may not affect a lot the working system of the sub-WSNs. In this general working model, a sensor node always acquires information from all the resources regardless of whether they are demanded or not. On the other hand, in some resource-sharing models, the provider or sub-WSNs can dynamically change the working period or node operations. These types of systems run on the principle of both query-answer and periodical data handling (e.g., ZigBee smart systems). In these systems, there is no need to obtain information from all the resources which a node has. Only the demanded resources and their functions are managed. The dynamic working models may be more efficient for interactive IoT WSN provider systems. However, in the dynamic working models, a large number of different clients can lead to unbalanced workloads for sub-WSNs and their resources. The proposed model in the paper considers provider systems that use the dynamic working model. The general working principle of an IoT WSN provider system is shown in Fig. I. The clients, who want to utilize the system, first pass the steps of registration and resource selection/allocation processes. Today the features of the WSN resources are generally advertised on the web through some internet technologies such as SensorML or other XML/JSON based applications [13]. Clients take advantages of the system by selecting the resources they want. During these processes, the general criteria are client parameters. On the other hand, these types of systems consist of sub-WSNs which are generally heterogeneous. In other words, the sub systems may have different types of features and operations. For example, a sub WSN sharing its own resources can run an owner-defined application. In another example, because of the demand difference, certain parts of a sub WSN can be busier than the other parts of the network.