Optimized Execution of Continuous Queries over Distributed Data:
A Cost-based Approach
=====================================================================

With the rapid development of network services, today a large variety
of dynamic data such as sensor values, stock prices etc. is available
to us at eye blink speed. With rapid availability has emerged the need
to harness the dynamic data and obtain answers to complex queries
(mainly aggregates) over this data. The issue of availability of
dynamic data has been studied in depth previously. A large number of
schemes have been proposed to organize effectively the repositories
inside a dynamic data dissemination network to make the data available
to the clients at greater speed and fidelity. However no significant
attempt has been made to develop a cost-based approach for answering
continuous queries over dynamic data. These queries are mainly associated
with an incoherency bound and thus can handle some degree of imprecision.
In this talk, I will discuss a novel approach to answer continuous
aggregate queries over dynamic data made available through a network
of data aggregators in a cost-effective way. The key behind this is to
identify the factors on which the overall cost for disseminating a
data item with a specific incoherency bound depends. Based on the cost
model, I will first discuss certain heuristics to effectively answer a
single sum query over the data items provided by the aggregators.
I will then explore techniques for efficiently handling multiple queries
posed by a client in the same data dissemination model.