Design and performance modelling of cloud-based sensing-as-a-service platform

Abstract

The area of sensing has witnessed signi cant development over the past two decades. Body of work on sensors has focused on individual sensors, sensor networks, in-network data processing and distributed query processing. Along with using dedicated sensors, smartphones are be- ing used for sensing, because of available on-board storage and computation capabilities. But gathering all the sensor data and o ering Cloud-based Sensing-as-a-Service(CSS) has not been explored. In a CSS platform, data from sensors is collected on a cloud-based server, then col- lected data is processed to provide services to the consumers. In CSS, the whole infrastructure of sensor data collection, followed by data processing to provide services to consumers, is o ered as a service on a cloud. CSS opens up considerable avenues for research pertaining to new and emerging commercial applications. Two folded contribution of this work can be described as below: 1. Architecture of Cloud-based Sensing-as-a-Service platform 2. Performance Modeling of designed platform The CSS platform consists of APIs for sensor data collection, APIs for using platform services, primitive queries, abstract queries and set of modules to process sensor data to provide consumer services. Platform services can be in terms of alert noti cations for consumer subscriptions or consumer initiated queries for particular type of event(s). The concept of abstract queries abstracts over underlying platform technical details, therefore allows end consumers to frame queries in user friendly manner. Each module processes one corresponding abstract query, which includes generating equivalent primitive queries and processing database results. Performance modeling, of a CSS platform, is about modeling usage of platform resources. Mod- elling is done in terms of incoming event rate and total number of subscriptions for platform services. A benchmark application, for measuring resource utilization, has been developed. The result of this benchmark application is resource usage data for complete duration of bench- marking. This result can further be used to come up with system design policy, about using a particular alert noti cation mechanism, for a CSS kind of platform.