Database Reference
In-Depth Information
generalize the two methods as snapshot monitoring and event monitoring. Let us
examine how the monitoring process works in each case.
Snapshot Monitoring
This method provides a snapshot or picture of the current
state of activities in the database system. Suppose you have a problem that is occur-
ring right now. Then you would want a snapshot of the database activities at this
moment. If a transaction executes for an inordinately long time, you may want to
take snapshots at short intervals to monitor how the transaction is progressing and
discover bottlenecks.
You may start the monitoring process and direct the system to take snapshots
at specified intervals. At these intervals, snapshots of performance variables are
recorded. You may define performance variables on the entire database, a specific
table, a disk space for tables, or other similar database objects.
Some systems let you gather cumulative numbers wherever applicable from the
start of the monitoring to the current snapshot. Snapshot monitoring systems can
also provide comparisons of performance variables at different snapshot points. The
comparisons may also be presented in the form of graphs that can display perfor-
mance trends, peaks, and valleys.
Typical snapshot monitoring systems have graphical interfaces that let you take
the following types of actions:
•
Set sampling frequency of performance snapshots
•
View results of performance computations
•
Define threshold values and alert actions
•
Generate alerts and view results of alert actions
Event Monitoring
With an event monitor, you capture information about
transient events you cannot normally monitor through snapshot monitoring.
An event monitor lets you monitor events such as deadlocks, completion of a
transaction, execution of a particular statement in application code, an insert to a
database table, and so on. In addition, you can also receive details about the
monitored event. You can know how long a transaction took to complete, how
much processor time was used by the transaction, and how many I/Os were
necessary.
The details about an event are stored in an event file that can be analyzed by
running an analyzer facility, usually as part of the event monitoring system. Details
generated by event monitors may be used to manage and tune database systems in
the following ways:
•
Analyze deadlocks, review locks in participated transactions, and correct lock
requests.
•
Monitor usage at application level and provide usage accounting.
•
Tune long-running transactions by monitoring operations such as reads and
writes within the transactions.
•
Analyze usage trends, predict future needs, and improve capacity planning.
•
Analyze data collected on usage of buffer pools, table areas, and tables.
Search WWH ::
Custom Search