Towards efficient monitoring in a sensor cloud

Sensor cloud is an extension of traditional sensor networks that virtualizes sensor nodes in order to achieve improved response time for client queries and reduction of energy consumption of physical sensor nodes in order to prolong their lifetime. In this paper we investigate information sharing in a sensor cloud with two types of sensors with different amount of data to send in each reading, and two types (classes) of client applications with different arrival rates. We show that caching of data readings can give noticeable improvements in performance, esp. at high arrival rate of client queries. Poll sharing amongst classes can also provide benefits but only in case where slower class polls are extended to update obsolete readings from the faster class (which may be done at no extra cost); other ways of sharing, including reading both values each time, are found to be inferior in terms of response time reduction. These findings can be operationalized to devise an efficient sensing policy for the sensor cloud.