The Caching tab on the Database Analysis drilldown graphically depicts the utilization of the various resource caches (memory areas) maintained for each database. These caches include package cache, catalog cache, the SQL workarea , and buffer pools—all of which enhance the performance of the database when their respective hit rates are high. Usually the hit rate is the ratio of retrieval requests that were satisfied by data already in the cache to the number of all requests—those satisfied by the cache plus those that required a disk read. (This ratio is expressed as a percentage in the various graphs on this tab.)
If you identify a cache utilization problem for a specific database on this tab, you can use the Caching tab on the Client Application Analysis drilldown to identify those applications that are contributing to the problem.
Use the following graphs to follow cache performance over consecutive monitoring intervals:
The Sorting tab on the Database Analysis drilldown uses graphs to show sort activity within the selected database and to contrast this activity to overall sorting activity in the instance.
An application query often requires that retrieved rows be put in a specific order—that is, be sorted—to satisfy a SQL sort request (as with the ORDER BY clause), to construct an index, or to perform a hash join. Sort heap is a block of memory that the database manager allocates to an application as workspace whenever the application requires a sort. An important factor in the performance of any sort is the size of its sort heap, which is determined by one of these parameters:
A sort consists of these steps:
The actual sort process—If the information being sorted cannot fit into the sort heap, some information is written to temporary database tables. This incident is called a sort overflow. Sorts that do not overflow perform better than those that do.
Return of the sorted rows—If sorted information can be returned directly without requiring a temporary table to store the final, sorted list of data, the sort is referred to as a piped sort. If the sorted information requires a table to temporarily store the data before it is returned, the sort is referred to as a non-piped sort. A piped sort always performs better than a non-piped sort.
Sorting problems, such as overflows, non-piped sorts, less than optimal sort heap allocation size, might indicate a need to adjust the SORTHEAP and SHEAPTHRES parameters.
Use the following graphs to determine sort activity and problems over consecutive monitoring intervals:
The I/O Activity tab on the Database Analysis drilldown provides a summary of I/O activity on the database or partition you are monitoring. The graphs track of rates for the following:
Reads and writes that involve the buffer pool
Direct reads and writes, which do not go through the buffer pool, but access the database directly
The I/O Activity tab contains the following graphs, which keep track of database I/O rates over consecutive monitoring intervals:
High rates depicted on the graphs for the following activities might be indications of performance degradation since these activities involve disk I/O and slow-downs in query processing:
Physical read rates (compared to logical read rates)
Synchronous read or write rates
Direct I/O rates for a database
If you need to investigate high rates for any of these activities, refer to the graphs on the I/O Activity tab in the following drilldowns:
Client Application Analysis drilldown—To identify the applications that are contributing to high I/O activity rates
Tablespaces drilldown—To pinpoint the tablespaces on which the high I/O activity is occurring
Buffer Pools drilldown—To identify those buffer pools that are not being used effectively to offset I/O on the database
The Locking tab on the Database Analysis drilldown gives you an overview of lock activity on the database. Scanning the graphs and statistics on this drilldown, you can quickly detect the following locking problems on the database:
High rates of lock holding or waiting
High numbers of deadlocks, escalations, or timeouts
Near-or full-capacity utilization of the lock list space.
If lock problems are evident, use the Client Applications drilldown to investigate the lock activity of individual applications to pinpoint those applications that are contributing to the locking issues.
Monitor the following graphs to detect abnormal lock activity on the database:
These lock activity counts provide more detail about any lock problems depicted in the graphs. For each activity type, you can view the count for the lifetime of the statistic (the number of occurrences since the database was activated or performance statistics were reset).
Additionally, the Statistics tab shows rates based on these counts. The names for these statistics begin with Lock.