1. 1 Towards a Flexible Global Sensing Infrastructure Chien-Liang Fok, Gruia-Catalin Roman, and Chenyang Lu

2. 2 SensorNets are evolving: Current software architecture does not easily facilitate –Integration of new technology –Application flexibility –Evolving global sensing infrastructure Small-scale, Homogeneous, Application-specific Problem Statement Large-scale, Heterogeneous, General Purpose Existing SensorNetsFuture global sensing infrastructure (GSI) evolving into

3. 3 Global Sensing Infrastructure A collection of sensors, microservers, base stations, and the Internet Heterogeneous and continuously evolving Shared by many users

4. 4 Two Example Applications Global supply chain monitoring –Sensors attached to products, cargo containers, loading docks, ships, etc. –Many different users Regional disaster scenario coordination –Highly heterogeneous and dynamic network –Involves many organizations and networks –Re-program GSI to help victims and coordinate responders

5. 5 Our Proposed Solution Create a new software architecture based on service-oriented computing & mobile scripts –Web services: global interoperability –Reactive platform-independent mobile scripts: dynamic application deployment in SensorNets –Efficient platform-dependent services: flexible service binding Declarative service-oriented programming paradigm enables seamless integration Simplify application development on evolving & heterogeneous networks

6. 6 Related Work Arch Rock Corporation [Woo, Sensys’06] –Expose SensorNet nodes to the external world as a web service using existing standards (WSDL) –We support mobile scripts (dynamic deployment of apps) for programming inside SensorNets SenseWeb [Santanche et al, IPSN’06] –A web portal for integrating & viewing data collected from a SensorNet IrisNet [Gibbons et al, IEEE Perv. Comp.’03] –A distributed database for WSN data Tenet [Gnawali et al, SenSys’06] –Support for tiered SensorNets –SensorNet has a fixed and uniform instruction set

7. 7 More Related Work Hourglass [Shneidman et al., Harvard TR’04] –An infrastructure for connecting SensorNets with the Internet –Focuses on data routing and processing ASVM [Levis et al, NSDI’05] –Create a VM with a custom instruction set –Does not provide dynamic service binding SOS + DVM [Balani et al, EmSoft’06] –A VM with dynamically loadable binary modules –Fixed binding between scripts and modules Melete [Yu et al, SenSys’06] –Multi-Application support within SensorNets –No quality of service provisions

8. 8 Our Software Architecture Web services –Global interoperability Execution engine (VM for mobile scripts) –Deployment & execution of mobile scripts Service repository –Local or distributed Service management framework –Discovery –Binding –Invocation

9. 9 Challenges: Programming Device Heterogeneity –Gracefully handle error conditions like when a script invokes a service that is not available Service Description Language –Declarative –Compact, extensible –Spatiotemporal –Heterogeneous Identify the boundary between scripts and services

10. 10 Challenges: Resource Management Quality of service for concurrent applications Varying resource availability –Processing capabilities –Memory –Network bandwidth –Sensors –Battery power

11. 11 Challenges: Runtime Coordination –Inter & intra application Event Distribution –Scripts must react to changes in their context –Wide range of event types, sources, priorities, and usage patterns Service Discovery –Limited resources prevent hosting every service locally –Distributed service repositories –Remote/local persistent/transient binding

12. 12 Current Status: Programming Service Description language Scripting language 1 require FireDetection 2 service serv; // a handle to the service 3 4 void main() { 5 bind(EAGER | PERSISTENT, THREE_HOP, serv, FireDetection) onError err(); 6 while (true) { 7 migrate(getRandomNeighbor(), STRONG); 8 boolean isFire = invoke(serv) onError err(); 9 } 10 }

13. 13 Current Status: System Extended Agilla’s VM to support mobile scripts & service invocation Integrated a service provisioning framework Tested on Tmote Sky testbed

14. 14 Conclusion Isolated SensorNets are evolving into a GSI A new software architecture is required for building flexible GSI applications Our proposed three-tiered service-oriented computing architecture: –Global web services –High-level platform-independent mobile scripts –Low-level platform-dependent services Many challenges remain