Abstract—Mobile devices such as smart phones are increas-ing permeating society. With strides in computational power, coupled with the ability to connect to other small devices, smart phones are able to host novel services. To address the repetitive problems associated with mobile service development, namely service reachability, scalability and availability, we have developed Odin, which is a middleware platform for mobile service provisioning. Beyond providing a provisioning solution, Odin conserves scarce resources such as network bandwidth and device power supply. However, Odin has previously lacked an ability to take into account operational context. In this paper, we present context-aware extensions to Odin that further optimise resource usage. Augmented with support for context types that include location, performance, power and network, Odin is able to propagate context information to applications and dynamic adapt the middleware’s behaviour. Novelty of the work lies in a solution whose device overhead is very low, and one that offers a coherent approach to context dis-semination and adaptation. Based on quantitative evaluation, context-aware Odin’s low overhead is demonstrated along with significant gains in resource conservation. I.

There has been growing interest recently in transforming small mobile devices from simple service consumers into service providers. The shift from consumer to provider has been made possible by hardware advances that enable small devices to host services. Combined with features like multihoming, small devices can be augmented with other devices to provide new services that exploit their inherent mobility. Intermediary-based middleware has emerged as a promising means for realising mobile services. Essentially, an intermediary is a host on a fixed network that contains a proxy for a mobile service, with clients consuming the service via the proxy. Our intermediary-based middleware ensures reachability and scalability of mobile services, and offers switching among heterogeneous network connections used to link services to their proxies. In this paper, we present further development, based on proxy migration, that enables optimisation of both device and intermediary-side resources, and which enhances mobile service availability.

Recent years have witnessed growing interest in using mobile devices to host services. The move from service consumer to service provider has been enabled by advances in mobile device hardware. Many contemporary devices also come equipped with multihoming, which allows a device to connect to heterogeneous networks, such as GPRS, WiFi and Bluetooth. To help leverage these new capabilities, we have developed middleware infrastructure that allows clients to consume services hosted on mobile devices. In this paper, we describe key functions of the middleware, namely ensuring reachability of roaming services and managing device connectivity in the presence of multihoming. Novel contributions of our work include session-preserving connectivity, proactive and reactive switching between network interfaces, and independence of network technologies and communication protocols. Proactive switching allows a device to initiate a switch from one network interface to another. Reactive switching responds to the loss of connectivity on one interface by changing to another. In both cases the logical connection, including any state, is preserved. Furthermore, logical connectivity is preserved when switching between heterogeneous networks and protocols. Our current implementation includes support for Bluetooth and HTTP/WiFi connectivity. Preliminary performance evaluation reveals that the middleware imposes little and acceptable overhead at run-time.

Dynamic reconfiguration is the process of making changes to software at run-time. The motivation for this is typically to facilitate adaptive systems which change their behavior in response to changes in their operating environment or to allow systems with a requirement for continuous service to evolve uninterrupted. To enable development of reconfigurable applications, we have developed OpenRec, a framework which comprises a reflective component model plus an open and extensible reconfiguration management infrastructure. Recently we have extended OpenRec to verify whether an intended (re)configuration would result in an application's structural constraints being satisfied. Consequently OpenRec can automatically veto proposed changes that would violate configuration constraints. This functionality has been realized by integrating OpenRec with the ALLOY Analyzer tool via a service-oriented architecture. ALLOY is a formal modelling notation which can be used to specify systems and associated constraints. In this paper, we present an overview of the OpenRec framework. In addition, we describe the application of ALLOY to modelling re-configurable component based systems and highlight some interesting experiences with integrating OpenRec and the ALLOY Analyzer