Data quality and query cost in pervasive sensing systems
Pervasive and Mobile Computing
| Hi-index | 0.00 |
Challenging issues of scale for networked servers arise from the number of clients that access each server. This thesis describes techniques for scaling data delivery from a shared-memory multiprocessor networked server to a large number of clients. We implement and explore connection-level parallelism for network protocols running on the server. We show that different implementations are appropriate for sending streaming media when compared with sending conventional data, such as web content. Matching the number of threads to the number of processors in the system yields the best overall performance when sending continuous media. Delivering the desired rate to each connection is accomplished by having threads directly measure their own performance and feed this information to the scheduler. For conventional data, matching the number of threads to the number of connections in the system yields the best aggregate throughput. In wireless sensor networks, challenging issues of scale arise from needing to satisfy a variable and potentially large number of concurrent queries for sensor data. In the second part of this thesis, we examine the benefits and costs of caching data for sensor network-based applications. We propose and evaluate several approaches to querying for, and then caching data in a sensor field data server. We show that for some application requirements (i.e., when delay drives quality-of-service), policies that emulate cache hits by computing and returning approximate values for sensor data yield a simultaneous quality improvement and cost savings. This win-win is because when system delay is sufficiently important, the benefit to both cost and quality achieved by using approximate values outweighs the negative impact on quality due to the approximation. In contrast, when data accuracy drives quality, a linear trade-off between cost and quality emerges. We identify caching and lookup policies for which the sensor field query rate is bounded when servicing an arbitrary workload of user queries. This upper bound is achieved by having multiple user queries share the cost of a sensor field query. Finally, we demonstrate that our results are robust to the manner in which the environment being monitored changes using two different sensor field models.