CPSC538A Paper presentation 3
The Design of an Acqusitional Query Processor for Sensor Networks
Paper Summary:
This paper introduced the idea of an acquisitional query processor (ACQP) for data collection to solve the where, when and how often data is physically acquired and delivered to query processor. It presented the query in SQL format for controlling data acquisition, and showed the related query optimization, dissemination and execution schemes. They designed and implemented an ACQP engine--TinyDB, and run it on each node in the sensor network. TinyDB runs on the Berkeley Mica mote platform on top of the TinyOS operating system.
First, it provided the properties of the sensor devices and the four phases of the power consumption. Based on that, it presented the SQL-liked queries (event-based queries, lifetime-based queries and power-based queries) in the consideration of the power consumption. Then, it discussed it's unique power sensitive dissemination and routing-semantic routing trees (SRT). SRT combined the attributes in the routing tree to make the query more efficient and disseminated less number of nodes. At last it discussed the prioritizing data delivery by using several queue schemes.
The motivation of the paper is in term of energy saving, which is very expensive in exchanging the power of sensor networks. All the relative experiments were done in the Lab to simulate the real world sensor network. The discussed ACQP is not very mature at this point. It still has to solve the difficulties in optimization of the queires.
Discussion Summary:
In lifetime-based queries, it used MIN SAMPLE RATE r. If the computed sample rate for the specified lifetime is greater than this rate, sampling proceeds at the computed rate; otherwise, use the fixed rate r. How to set the minimal sample rate?
In order to optimize the queries, it need to know the cost of each kind query. Then do the statistic to get the ordering. The question remains that is how to get those kind of statistic properties, is that feasible in the real-world sensor network?