Database Reference
In-Depth Information
example, if you want to record the number of active transactions every 20 minutes,
this method will produce the numbers three times every hour. The sampling method
has minimal impact on system overhead. In the event-driven method, recording of
statistics happens only when a specified event takes place, not at regular intervals.
If you want to monitor growth of an index table, you can set the monitoring mech-
anism to record the event every time an update is made to the index table. We will
elaborate on these methods a little later. First, you must appreciate the purpose of
database monitoring and what monitoring is expected to accomplish.
Purpose of Monitoring
After deployment, actual use of the new database system reveals unanticipated dif-
ficulties and factors. During implementation, the project team plans according to
the expected usage and access patterns. Not even extensive testing could expose all
the variations. Only after the database is sufficiently exercised by the users can the
DBA gain real insights.
Actual use for a reasonable period, with execution of all types of transactions,
exposes bottlenecks affecting performance. Perhaps the size of the memory buffers
constricts the flow of transactions. It is possible that the initial index plan for a
particular table hinders quick access instead of speeding up the transactions. All
such bottlenecks surface as the database is used more and more. Therefore, ongoing
monitoring and collection of statistics becomes important.
At a high level, two major purposes prompt database monitoring and collection
of statistics:
•
Discerning the growth pattern, estimating additional resources, and planning
for future needs
•
Tuning the database system for performance
At a more detailed level, database monitoring enables the DBA and other con-
cerned members of the project team to accomplish the following:
•
Identify and analyze performance problems in the database system configura-
tion or in the applications using the database system.
•
Utilize early warning systems to be alerted to potential problems.
•
Discover exception conditions based on defined threshold values.
•
Wherever possible,
automate corrective actions in response to detected
problems.
•
Define additional requirements of statistics to be produced in addition to
system defaults.
•
As and when changes are made to the database system after deployment, assess
the effects of the changes.
The Monitoring Process
Above we briefly mentioned two common methods for monitoring the performance
and the state of the database system. Now let us get into more details. Let us
Search WWH ::
Custom Search