Author :
Schiele, Gregor ; Ranganathan, Anand ; Campbell, Roy ; Paluska, Justin Mazzola
Abstract :
Context-awareness, dynamism, and heterogeneity are some of the properties that differentiate pervasive computing from traditional distributed systems. Most traditional distributed systems are unaware of context, are static, and are composed of homogeneous devices. As a result, the assumptions underlying traditional middleware infrastructures differ from the ones for pervasive computing. In a pervasive computing environment, issues such as mobility, disconnection, and dynamic introduction and removal of devices and merging of the physical environment with the computational infrastructure are common and affect the underlying middleware infrastructure. Furthermore, different devices might be connected to different networks, with different latency and bandwidth. As a result, the middleware must provide mechanisms for handling disconnection, addressing fault tolerance, and adapting to a number of issues related to diversity including heterogeneous device resources. The scale of pervasive computing in terms of the number of devices and services, combined with the lack of a single system administrator, the associated dynamism, and frequent failures requires middleware services capable of evolving and reorganizing themselves.