Gossip Algorithms in WSN with Random or Selective Choice of Neighbor Nodes based on Update Status Blerina Zanaj Department of Informatics and Mathematics Agricultural University of Tirana Albania Elma Zanaj Departaments of Electronics and Telecomunications Polytecnic University of Tirana Albania Abstract—Epidemic algorithm applied in Wireless Sensor Network (WSN) is classified as a diffusive type of algorithm due to the way the information is spread in network nodes [1,2,3]. Gossip algorithms as well belong to this class of algorithms and the interest on them is increasing every day more because they offer an efficient way of information transferring through the wide spread systems. It is very important to model their behaviors in order to get an insight view of their performance and of the application built based on them. In this work is tested the performance of two Gossip algorithms through simulations [3,4]. It takes into account the presence of additional packets in the network, the total time spent for the whole system updating and the energy consumed for transmission /receiving of the different kind of packets exchanged between the nodes. Keywords—Soread Systenm, Knowledge Of Updating Status, Failed Node, Total Updating Time Of System. I. INTRODUCTION The increasing interest in WSN is due to their operability as a network compound of a large number of sensors that collect information from the surrounding environment for different parameters. They are used in many applications but at the same time they face a lot of challenges like saving energy, node localization during mobility. For solving this problem different algorithms are developed which help to localize the nodes but they also find the best path with the less possible energy consumption during the communication and information exchange between the nodes. The implementation of those techniques and algorithms in a network there are done with methods called “Plug and Play” [1,2]. The method had made it possible to sensor to be cheap and to have many different applications and functions by changing only their software without intervention into their hardware. During this study will be described Gossip algorithms that are simple to implement. They can fast adapt with size growth of the network. They also have good stability in the presence of different fields of failure like in the process and in links. In Gossip algorithm every node will send data to a random chosen node. In the moment when the destination node gets the packet, it is its turn to choose a random node to send the information. The procedure continues infinitely till the packets arrive in the destination or to the farthest point of network distinguished by the maximal number of hops. In this work are shown two Gossip algorithms that Random (RG) and With Knowledge of Update Status of Neighbor (GKUS) node. These algorithms deliver information through the network in a way to reduce the number of exchanges done in the network. The performance is compared in the terms of total time needed to update the network, extra exchanges and communication and energy consumption [4]. In simulation model is supposed that only one node is the initiator hold the updated information. The initiator is responsible for the starting of updates so that even other nodes take the information. It is supposed that all the nodes have an identification unique number. In RG a node chooses another node randomly and it decides to update it. All the nodes recognize the neighbor nodes around them but the nodes do not have any clue about their update status. In this situation we expect to have excess messages. GKUS, the nodes learn about the update status of the neighbor node before they send the information. This is achieved by sending first a request for the nodes status and depending on the response returned is overtaken a concrete action, is sent yes or not an update. When a node is ready to send information, it chooses randomly a destination node like in the case of RG. To the chosen node is first sent a request packet when is asked to respond back with a packet notifying its updating status. If the response from the destination node is positive meaning that yes it is already updated, then the process starts over again with the choosing of another destination node and again a request process restarts. If the answer from the destination node is a no than the source node sends the updating information. In meantime both the nodes addresses are removed from the respective lists of the neighbor nodes that each node contains. This process of random choosing of nodes continues till the list of neighbor nodes becomes empty. As each node is updated only once there are no excess packets but there are information messages in excess. By reducing the most possible the request/confirmation packets and to a fix size is achieved a good performance with this algorithm [3,4]. The paper is organized as follows: Section II presents the system parameters setting important for simulation and results obtained in the work. The algorithm description is shown in section III; in section IV are brought the simulations done testing different parameters; section V brings a summary of the conclusions discussed so far in the work and in the last section number VI are written the references. International Journal of Engineering Research & Technology (IJERT) ISSN: 2278-0181 http://www.ijert.org IJERTV5IS060024 (This work is licensed under a Creative Commons Attribution 4.0 International License.) Published by : Vol. 5 Issue 06, June-2016 www.ijert.org 40