Database Reference
In-Depth Information
condition is done in the optimization time but its
evaluation is delayed until the execution time.
Thus, the query is guaranteed to be contained in
the view if the guard condition evaluates to true.
The query plan must also contain an alternative
subplan, called a fallback plan that computes the
query expression from other input sources in case
the guard condition evaluates to false.
For the view maintenance, as only a small
number of row are actually materialized so the
view can be maintained more efficiently. However
the existence of an exist clause in the dynamic
view syntax prevent to apply the incremental view
maintenance algorithms. The authors discuss how
to update the view incrementally by replacing this
clause. This can be achieved by either converting
the subquery to a join clause or converting the
view into an aggregation view and this depends
on the control table if it contains duplicated tuples
or not. However, this cache policy, which admits
the new value on the second access, can cause an
overhead and reduce the global performance of
the system. For example if the workload is noisy,
the system requires strong evidence on the pay-
off of switching between the materialized rows
due to short-lived transitions in the workload.
While CoLT prevents the Self Organizer from
“thrashing” when the workload is noisy by to
selecting an index for materialization based on
its long-term potential.
In conclusion, unlike most of previous work,
our approach is dynamic and provides a replace-
ment policy.
(2) New PCs can be added easily at any
time in order to meet the company's new
requirements.
(3) A centralized data server can become a
bottleneck as a large user community con-
tends for data server access.
(4) Data is unavailable when a failure occurs
on the network. In these contexts, users are
spread over the network, and each location
may have different types of query charac-
teristics and/or performance requirements.
The view selection problem and its generaliza-
tions will play an even greater role in these contents
where data need to be placed intelligently over
a wide area network. As the scale of distributed
systems and applications keep growing, the peer to
peer communication paradigm is now recognized
as the key to scalability. In a peer to peer system,
each peer may act both as a client and server so
that the number of servers increases linearly with
the size of the system thus ensures the scalability.
Our focus is on peer-to-peer distributed database
system (PDBS). Peers in a peer group have their
own local databases but can also retrieve data at
different rate from various points on the network.
Information in these local databases can be shared
among peers through user queries. PDBS can
obtain tremendous performance and availability
benefit by employing materialized views. A key
factor to ensure good performance in such context
is intelligent placement and replication of data
at different nodes on the network. However, the
membership and the query workload at any peer
are dynamic and unpredictable therefore the pro-
posed solution should be dynamic and treats the
queries as it arises at any peer. We argue that the
materialization technique can be applied in a P2P
DBMS context for (i) saving work on frequently
asked queries and (ii) increasing availability in
cases of failure.
FUTURE TRENDS
Recently, businesses are beginning to rely on
distributed rather than centralized database for
many reasons:
(1) PC processors are cheaper and significantly
more powerful than one big mainframe
computer.
Search WWH ::
Custom Search