PhD Qualifying Examination


"Adaptive Query Processing in Data Stream Management Systems"

Mr. Yin Yang


Abstract:

In this survey, we review adaptive query processing techniques in data
stream processing systems (DSMS). In a typical DSMS model, users register
long-running continuous queries, which are re-evaluated after the arrival
of each stream element. Because various stream characteristics may change
in an unexpected manner, an adaptive approach to query execution is
crucial for performance: as data characteristics change, execution plan
automatically change as well.

Adaptive query processing techniques can be classified into three
different categories. The first category of approaches, called static plan
based approaches, use traditional query optimization techniques to
generate static execution plans. When an environmental change happens, the
query engine generates a new execution plan and dynamically migrate the
query processing to the new plan. Because the cost of migration is
typically high, these approaches are suitable when data characteristics
are relatively stable.

The second category of approaches run queries based on novel execution
plans. These plans are different from traditional execution plans in that
these plans can adjust themselves according to changing environments. The
plans bearing this property are usually not as efficient as traditional
ones and some of these approaches are only discussed for specific user
queries.

Finally, the last category of methods, notably the Eddies approach, avoid
using query execution plans altogether and adopt a routing mechanism
instead. The routing mechanism is continuously adapting to the
environment. These approaches are suitable for extremely fluctuating
environments.


Date:     		Monday, 16 January 2006

Time:                   10:00a.m.-12:00noon

Venue:                  Room 2504
			lifts 25-26

Committee Members:      Dr. Dimitris Papadias (Supervisor)
			Dr. Qian Zhang (Chairperson)
			Dr. Lei Chen
			Dr. Mordecai Golin


**** ALL are Welcome ****