The Impact of Network Topology on Collection Performance

Opportunistic, Receiver-Initiated Data-Collection Protocol

“…The strengths of CTP are its ability to quickly discover and repair path inconsistencies and its adaptive beaconing, which reduces protocol overhead and allows for low radio duty cycles. Arbutus [15] is a protocol similar to CTP, but promises higher performance in weakly connected networks with poor node connectivity. Unlike CTP and Arbutus, ORiNoCo avoids control packets by adding minimal information to the beacons produced by the underlying low-power probing MAC protocol.…”

“…Low-power medium access protocols were designed to buy extended network lifetime at the cost of increased latency [12,20,5]. These techniques were used to devise both generic [1,7,15] and special purpose [3,10,13] collection protocols. Despite the success of these protocols, they impose some extra cost on the network by requiring up-to-date knowledge of state information; particularly neighborhood tables have to be exchanged-including node IDs and link-quality metrics (e.g., LQI, RSSI, or packet success rates).…”

Characterization of the impact of resource availability on opportunistic computing

“…A distributed scheme for calculating the level of resources available in the vicinity of a node in a delay-tolerant network is presented in [18], with a specific focus on buffer space, energy, and bandwidth. Our work in this paper leverages our previous work on the decoupling of the impact of the network topology from the one of protocol performance [10], where the impact of network topology is distilled down to one single metric, the Expected Network Delivery (END). The END metric is able to determine when the topological conditions of the network represent a bottleneck that throttles the achievable performance of a data collection protocol.…”

“…The END was developed for low-power wireless, but its applicability goes way beyond the world of low-end radios, and indeed it forms the basis for the work in this paper, where connectivity is generalized to resource availability: in SCAMPI's OC networked settings, it is no longer enough to be connected, but services and resources are to be shared across nodes. This paper attempts to adapt the logic reasoning that lead to the END in [10] to SCAMPI's heterogeneous networks to capture the impact of a generalized topology of services and resources on the performance of OC schemes.…”

“…Shared services must in fact be supported in a network environment that is made challenging by topological instability and constrained resources. In our previous work, we have investigated topological instability [10] and limited resources [11] using low-power wireless as a case study. Low-power wireless represents an exceptional test bench because it contains an exacerbated version of the most adversarial features of an OC setting (network instability and constrained resources) while offering easily accessible large-scale experimental resources [12,13,14] whose challenges [15] more than make up for the lack of mobility.…”