Measured Characteristics of FutureGrid Clouds for Scalable Collaborative Sensor-Centric Grid Applications Geoffrey C. Fox School of Informatics and Computing and Community Grids Laboratory Indiana University, Bloomington IN 47408 USA gcf@indiana.edu Alex Ho, Eddy Chan Anabas, Inc., 580 California Street, Suite 1600, San Francisco, CA 94104, USA {alex.ho, research.eddychan}@anabas.com ABSTRACT The emergence of cloud technology has raised a renewed emphasis on the issue of scalable on-demand computing. Cloud back-end support of small devices such as sensors and mobile phones is one important application. We report our preliminary study of measured characteristics of distributed cloud computing infrastructure for collaboration sensor-centric applications on the FutureGrid [1, 2]. We focus on understanding the characteristics of the underlying network and its impact on multipoint, distributed cloud scalability. We report our findings in areas of performance, scalability and reliability at the network level using standard network performance tools. We measure data at the message level using the NaradaBrokering system [3-8] by the Indiana University Community Grids Laboratory which supports a large number of practical communication protocols. Results are also presented at the collaboration and communication applications level using the Anabas sensor-centric grid framework [9], a message-based sensor service management and sensor-centric application development framework. Geographically distributed and heterogeneous clouds in the FutureGrid are used because of their support for scalable simulations. Our preliminary data indicates that a heterogeneous cloud infrastructure like FutureGrid coupled with a flexible collaborative sensor-centric grid framework is suitable for the study and development of new, scalable, collaborative sensor-centric system software and applications. KEYWORDS: distributed cloud, heterogeneous cloud, collaboration, sensor-centric applications, scalability, FutureGrid 1. INTRODUCTION Cloud computing services promise infrastructure resources to support application scalability. There are ample studies with systematic evaluation of this emerging information technology infrastructure [10-19] but few are on collaboration applications in general. There is even fewer work on leveraging heterogeneous clouds for real- time, distributed, collaborative sensor-centric applications in particular. Increased use of collaborative sensing in a wide range of social, environmental, commercial and military types of applications is being driven by the need for better information about the environment or operational picture of interest and the advancement of technology which provides smaller, inexpensive and more capable sensors. For instances, some collaborative sensor-centric applications could be found in the fields of environmental monitoring, security surveillance, or target tracking [20]. One example of an interesting application is the sharing of filtered, neighborhood parking meter sensor information regarding available parking spots via local street signs to smartphones [21]. In recent years, technology has enabled a noticeable shifting from using few expensive and feature-riched sensors to deploying a large number of small, inexpensive commodity sensors with some level of direct or indirect networking capability. This technology trend should continue for the foreseeable future. Therefore, there will be a growing demand for scalable support of collaborative sensor-centric applications that could utilize a wide variety of sensor types and a massive number of globally deployed sensors for timely and actionable decision- support scenarios.