Previous Topic: Brief Conclusion Text ElementNext Topic: Boolean Data Cell


Data Cell Types and Use

The Performance Manager sub-record types in the following table are stored in Performance Manager data records. You can access the fields within the sub-records as data cells within a rule construct. See Appendix C for the names of the data cells. You can also dump them using the ADVISE COLLECT REPORT DUMP_DATACELLS command. This command produces a dump report of all data cells in the LOCAL domain.

The following table shows the data sub-record types and associated domains:

Sub-record Type

Associated Domain

Communication

COMMUNICATION

Configuration

CONFIGURATION

CPU

CPU

Disk

DISK

Hot_Files

FILE

Metrics

COMMUNICATION, CONFIGURATION, CPU, DISK, FILE, LOCAL, PROCESS, TAPE

Parameters

COMMUNICATION, CONFIGURATION, CPU, DISK, FILE, LOCAL, SUMMARY, PROCESS, TAPE

Process

PROCESS

Tape

TAPE

Note: The CLUSTER domain references a sub-record type that exists only in memory. It does not exist in the Performance Manager data records.

If you write a rule in the LOCAL domain, the data cells available to the rule conditions and rule evidence include all values from the metrics and parameters records for a single node. The Performance Manager evaluates the rule once for each interval record (2 minutes for daily data, user-defined for history data), for each node that is processed. For example, an hour of daily data at the default interval contains 30 records per node for the LOCAL domain.

If you write a rule in the PROCESS domain, data cells available to the rule include all values from the process sub-record. Metric and parameter sub-record data cells are also available from the PROCESS domain because they are part of the current interval record containing the given process sub-record for the current node. The Performance Manager passes each of the process sub-records through all rules in the PROCESS domain for each interval record and for each node.

The number of process sub-records that the Performance Manager examines depends on the number of processes logged in, the time range, system activity, and for history files, the workload classification scheme. You must reference data in a process sub-record from the PROCESS domain or from the LOCAL, COMMUNICATION, CONFIGURATION, CPU, DISK, FILE, or TAPE domains with a process domain index specifier. The Index Specifier Data Cell section describes Index Specifier data cells.

The SUMMARY and CLUSTER domains do not directly correlate to Performance Manager data records; however, they are derived from them.

The data cells available in the SUMMARY domain are metrics that are maximums or averaged from all of the interval records. After processing all the data for a single node, the summary data cells become available in the SUMMARY domain. The last parameter sub-record is also available in the SUMMARY domain. The number of times the Performance Manager tests a rule in the SUMMARY domain is equal to the number of nodes processed.

After the Performance Manager processes all nodes data, the data cells available in the CLUSTER domain are disk statistics that represent the cluster perspective of the I/O traffic. The Performance Manager combines data from each node's two-minute disk sub-records into a set of two-minute cluster records in memory. The Performance Manager provides items such as cluster-wide throughput and operation rates; however, there is no longer any association with the “current two-minute” data record. Metric and parameter sub-record values (available from the LOCAL domain) are not accessible from the CLUSTER domain.

The following table lists the seven types of data cells:

Data Cell Types

Brief Descriptions

Boolean

1.0 (True) or 0 (False)

Numeric

Floating point value

String

An ASCII string

Time

Clock time for data

Scan routines

Loads tally data cells and returns sub-record count

Tally

Derived data

Index specifier

Index to a sub-record