| Since few years we observed that number of computing devices increased quickly and found application in all domains and also among individuals. From car to cellphone, including bathroom scale, computing in all its forms and variations is ubiquitous in everyday life, and continues to expand its scope of action.At the crossroads, there are wireless sensor networks. These networks are composed of many small autonomous devices, which can self-organize them to communicate with each other and work together to capture information about their environment, without any human intervention. These devices are inexpensive, but inefficient, and can be deployed in mass without taking care of each one. They are the current ultimate stuff of mass computing deployment and absolute need of networks.However in current traditional architecture, these devices are not really independent since they require the presence of sink and gateways in order to use them with present infrastructure.In this paper we are committed to presenting an architecture that allows overcoming these constraints that undermine the concept of total independence of devices, and more, which can damage the overall performance of sensor networks, in particular in specific situations. For that we have proposed architecture based on a peer to payer overlay which will allow us to collect data from the sensor network. This collect can be done either through a history of recent values, which can limit the network load by partially centralizing data, either through a direct connection between an end user and a sensor.With this expansion, computer networks, especially wireless networks are also growing. It seems far the time where we have phone calls only from house or from phone booth, nowadays everybody wants to be connected at any time. And beyond people, now more and more devices are connected by themselves, without any user interaction.We also performed a partial implementation of this architecture and made some tests as a proof of concept of our architecture. If these tests cannot provide a definitive conclusion on the performance of our architecture, they have allowed us to conclude on the feasibility of such architecture and that there is not an obvious risk of widespread failure. |
