Cube
The Cube menu shows the load distribution over a long period in a single hitmap and lets you pick the interval you are interested in and drill down into detailed metrics. It is provided only in the Oracle Pro product.
To track a performance issue, you must first find when it occurred. If you widen the query range of a metric chart to a full day, short spikes are flattened out and go unnoticed. If you narrow the range, you cannot tell which interval to look at. Cube divides the entire query period into 5-minute or 1-day cells and displays them by color intensity. You can locate the interval where load is concentrated by color first, then analyze only that interval in detail.
On this screen, you can do the following:
-
Select a metric and check its distribution over the entire query period on a hitmap
-
Select a cell with high load on the hitmap and check the OS, session, stat, and wait metrics for that interval
-
Check the Top SQL for the same interval
Viewing the hitmap
Specify the query conditions for the hitmap in the left area of the screen.
-
In Start date, select the date to start the query from.
-
In Query period, select the unit of the hitmap. Day (5 min) displays a day in 5-minute cells, and Month (daily) displays a month in 1-day cells.
-
In Aggregation, select the method to apply when a single cell contains multiple values. You can select Average or Max.
-
In Metrics, select the metric to display on the hitmap.
-
In the Instance selector at the top right of the screen, select the instance to query.
Metrics
You can select any metric in the db_oracle_dma_counter category. The most commonly used metrics are as follows.
| Item | Metric | Description |
|---|---|---|
| CPU | cpu(xos) | Server CPU usage collected by the XOS agent |
| Memory | mem(xos) | Server memory usage collected by the XOS agent |
| Active Sessions | active_sessions | Number of active sessions |
| DB Time | DB time | Time the database spent processing requests |
| Logical I/O | session logical reads | Number of session logical reads |
| Physical I/O | physical reads | Number of direct reads from disk |
| SQL Executions | execute count | Number of SQL executions |
| Lock Wait | lock_wait_sessions | Number of sessions waiting for a lock |
Analyzing a detailed interval
When you select a cell on the hitmap, the detailed metrics for that interval are displayed on the right side of the screen. If you query by Day (5 min), the 5-minute interval of the selected cell becomes the analysis target.
The tabs in the detail area are as follows.
| Item | Description |
|---|---|
| OS | Trends for CPU, memory, disk I/O, network I/O, and CPU I/O wait collected by the XOS agent |
| Session | Trends for active sessions, connections, DB Time, lock wait sessions, long-running sessions, and enqueue waits |
| Stat | Trends for SQL executions, transactions (commit and rollback), logical I/O, physical I/O, block changes, and redo |
| Wait | Wait event distribution and latch and enqueue metrics for the selected interval |
| Top-N | Top SQL list for the selected interval |
The Wait and Top-N tabs use the same widgets as the Wait analysis and Top SQL screens. For more information about the two screens, see Wait analysis and Top SQL.
The Top-N tab includes the option bar and the analysis criteria tabs of the Top SQL screen as they are. The interval selected in Cube is applied as the query time.
In the Session tab, you can check the active session list and the process information for the selected interval. Because this screen queries a past interval, the session list is not refreshed in real time. The active session and process information are each available on their own subtab.