Introduction
Efficient
support for continuous queries is critical in many application
areas, including sensors networks, traffic monitoring and intrus
ion
detection. The data-intensive nature of these applications, and
the fact that they often compare and combine the results of quer
ies
on data arriving on the wire with the results of queries on info
rmation
stored in the database provides the strong rationale for current
research projects that focus on adapting and extending SQL and
its enabling technology to work on data streams. Unfortunately,
SQL is facing severe limitations in this new role, which will
severely impair the power and generality of SQL-based management
systems. A main objective of Stream Mill is to overcome these
limitations, and thus achieve a much broader range of usability
and effectiveness in its application domain than other stream
management systems. In particular, Stream Mill will address the
following issues:
- support
for typical sequences queries, e.g. to search for patterns,
which SQL does not support
- the loss
of expressive power due to the fact that blocking operators
are no-longer usable on data streams--- further reducing the
limited expressive power of SQL, and
- many
applications use streaming XML data, rather than relational
data streams.
We are developing the enabling technology to overcome these proble
ms.
In particular, Stream Mill unifies the processing of relational
streams and XML streams--much in the same way in which research
prototypes and commercial DBMSs are now moving to unify the manage
ment
of relational data and XML data. Data Streams will also support
queries on sequences and ordered data, since they are needed on
data streams as much as they are on stored sequences. Finally, Str
eam
Mill will compensate for the loss of expressive power caused by
blocking query operators, through powerful new operators that are
nonblocking, such as the continuous UDAs of ATLaS.
We also exploring techniques for replacing blocking operators with
nonblocking ones when these are used in expressing blocking querie
s.
|