6.1.3 Exadata Metrics

Exadata metrics are recorded observations of important properties or values relating to the Exadata system software.

Exadata metrics contain detailed statistics for most Exadata components. Most metrics relate to the storage server and its components, such as flash cache, cell disks, and grid disks. Storage server metrics enable detailed monitoring of Exadata storage server performance.

Metrics are of the following types:

• Cumulative metrics are statistics that accumulate over time since the metric was created or the server was restarted.

  For example, CL_IO_RQ_NODATA is the total number of I/O requests that didn't return data.

• Instantaneous metrics contain the current value at the time of the metric observation.

  For example, CL_MEMUT is the percentage of total physical memory currently used on the server.

• Rate metrics are computed statistics where the value is observed over time.

  For example, N_NIC_KB_TRANS_SEC is the number of kilobytes per second transmitted over the server's Ethernet interfaces.

Some metrics differentiate between small I/O and large I/O. For such metrics, small I/O means I/O that is less than or equal to 128 KB in size. Large I/O is greater than 128 KB in size.

By default, metric collections occur at 1-minute intervals. Metric observations in the default collection are initially recorded in memory and later written to a disk-based repository. For most metrics, historical observations are maintained for seven days by default. However, a subset of key metrics are retained for up to one year. Starting with Oracle Exadata System Software 24.1.0, you can view and control which metrics to maintain for up to one year. In all cases, Exadata automatically purges historical metric observations if the storage server detects a shortage of space for the disk-based metric history repository.

Commencing with Oracle Exadata System Software 22.1.0, you can optionally configure fine-grained metrics. To enable fine-grained metrics, you must specify a collection interval between 1 and 60 seconds. You can also choose the metrics to include in the fine-grained collection. Fine-grained metric collection is the foundation for real-time metric streaming. Consequently, fine-grained metrics are only recorded in memory to support efficient streaming to an external metric collection and visualization platform.

You can use the CellCLI LIST command as the primary means to display Exadata metric definitions and observations. See Using Metrics. However, because Exadata metrics are managed within each Exadata server, metric observations must be collected and correlated by some additional means to gain a system-wide view.