A programming model and system support for disconnected-aware applications on resource-constrained devices

Support for disconnected operation via architectural self-reconfiguration

“…On the one hand, a number of promising solutions to the disconnected operation problem have already emerged. On the other hand, these solutions have focused on specific system aspects (e.g., data caching, hoarding, and replication; special purpose, disconnection-aware code) and operational scenarios (e.g., anticipated disconnection [15]), often requiring significant involvement by the system's (human) operators. While each of these solutions may play a role in the emerging world of highly distributed, mobile, resource constrained environments, our research is guided by the observation that, in these environments, a key determinant of the system's ability to effectively deal with network disconnections is its deployment architecture.…”

“…• Caching -locally storing the accessed remote data in anticipation that it will be needed again [8]; • Hoarding -prefetching the likely needed remote data prior to disconnection [9]; • Queueing remote interactions -buffering remote, nonblocking requests and responses during disconnection and exchanging them upon reconnection [7]; • Replication and replica reconciliation -synchronizing the changes made during disconnection to different local copies of the same component [8]; and • Multi-modal components -implementing separate subcomponents to be used during disconnection [15]. None of these techniques changes the system's deployment architecture.…”

“…None of these techniques changes the system's deployment architecture. Instead, they strive to improve the system's availability by sacrificing either correctness (in the case of replication) or service delivery time (queueing), or by requiring implementation-level changes to the existing application's code [15].…”

“…For instance, unlike [15] our approach does not require any recoding of the system's existing functionality or human intervention; unlike [8] it does not sacrifice the correctness of computations; finally, in comparison to [7] it minimizes service delivery delays. We directly leverage a software system's architecture in accomplishing this task.…”

“…For this reason, highly distributed and mobile systems are challenged by the problem of disconnected operation [15], where the system must continue functioning in the temporary absence of network connectivity. Disconnected operation forces subsystems executing on each network host to temporarily operate independently from other (disconnected) hosts.…”