Energy-Efficient Data Management for Sensor Networks
Dr. Niki Trigoni, Cornell University
Wednesday, January 21, 2004
4:30pm - SENSQ 5317
Joint Pitt/CMU database group meeting
Abstract
A powerful paradigm in sensor network design has emerged recently in which clients "program" the sensors through queries in a high level declarative language (such as a variant of SQL). Considering that sensors have strong constraints on their energy usage, this database approach to sensor networks calls for energy-efficient query processing and data dissemination techniques. A significant amount of energy can be preserved by (1) designing efficient plans for query execution and (2) by carefully coordinating data and query transmissions across the network.
The first part of my talk describes work on opportunities for multi-query optimization in sensor networks. In order to minimize the number of messages for a given query and data generation workload, a hybrid pull-push model is proposed, in which relevant data is collected at sensor nodes and pushed to "view" nodes, from where the data can be pulled when queries are issued. The goal is to decide, given a query workload, what data to store and where in the network, in order to minimize the expected overall query cost. Pull-push data dissemination is combined with powerful multi-query optimization methods for aggregate queries, resulting in a set of distributed algorithms that greatly reduce energy usage.
The second part of my talk focuses on scheduling nodes and coordinating their transmissions such that data flows quickly from event sources to storage (or query initiation) nodes while avoiding collisions at the MAC layer. Since all nodes adhere to the schedule, most nodes can be turned off and only wake up during well-defined time intervals, resulting in significant energy savings. Routing protocols can be modified to interact symbiotically with the scheduling decisions, resulting in significant energy savings at the cost of higher latency.
Bio Sketch
Niki Trigoni obtained a B.Sc. in Computer Science from the Athens University of Economics and Business (Greece) in 1998, whilst working as a data analyst in the I.T. department of the National Bank of Greece (1995-1998). Having won a National Scholarship from the Greek State Scholarhip Foundation (I.K.Y.), she obtained her PhD from the Computer Laboratory of the University of Cambridge in 2001. Her dissertation entitled "Semantic Optimization of OQL Queries" established a rigorous basis for the Object Query Language (OQL) proposed by the Object Data Management Group. This involved research in type checking, type inference, translation of queries into calculus and algebraic representations, syntactic and semantic optimization and physical query execution. The emphasis was on the use of association rules with exceptions for the transformation of queries into more efficient forms.
Dr. Trigoni is currently pursuing a post-doctoral fellowship at the Cornell Department of Computer Science under the guidance of Professor Johannes Gehrke. Her present research interests are in the areas of query processing and distributed data management in sensor networks. The finite battery power of sensors and the limited bandwidth of the wireless network impose severe constraints on query execution. The main goal is to build an energy-efficient data management layer for sensor networks by carefully selecting where to store data in the network, coordinating edge activations and devising multi-query optimization techniques.