7.3.4 The rpd_ml_stats Table

Note

The Performance Schema table described here is available as of MySQL 9.0.0. Support for HeatWave GenAI routines is available as of MySQL 9.0.1-u1.

The rpd_ml_stats table tracks the usage of successful HeatWave routines. These metrics reset whenever the respective DB system restarts.

The following HeatWave AutoML routines are tracked:

• ML_TRAIN

• ML_EXPLAIN

• ML_PREDICT_ROW

• ML_PREDICT_TABLE

• ML_EXPLAIN_ROW

• ML_EXPLAIN_TABLE

The following HeatWave GenAI routines are tracked:

• ML_GENERATE

• ML_EMBED_ROW

The rpd_ml_stats table has these columns:

• STATUS_NAME

  Identifies the type of meter tracking usage.

• STATUS_VALUE

  Displays metrics for metering. Content is displayed in JSON format.

Metrics in the table are entries as JSON values. The following metrics are used:

• n_cells

  The total number of table cells processed by the HeatWave AutoML routine for all invocations.

• n_cells_user_excluded

  The total number of table cells manually excluded for the HeatWave AutoML routine.

• n_blob_cells

  The total number of table BLOB cells processed by the HeatWave AutoML routine for all invocations.

• table_size_bytes

  The total number of bytes of data processed by the HeatWave AutoML routine for all invocations.

• blob_size_bytes

  The total number of bytes of BLOB/TEXT data processed by the HeatWave AutoML routine for all invocations.

• model_size_bytes

  The total number of bytes of data for the HeatWave AutoML model that is trained. This includes any explainer models. This metric only applies to the ML_TRAIN and ML_EXPLAIN HeatWave AutoML routines. All other routines will display NULL values.

• input_size_bytes

  The cumulative size in bytes of all input string/document invocations ingested by the HeatWave GenAI routine.

• context_size_bytes

  The size in bytes of the context string referenced when generating the response. This metric only applies to the ML_GENERATE HeatWave GenAI routine since the ML_EMBED_ROW routine does not have context. The metric will still appear for ML_EMBED_ROW, but will display a value of 0.

• output_size_bytes

  The cumulative size in bytes of responses generated by all invocations for the HeatWave GenAI routine.

• n_invocations

  The total number of times the routine has been successfully invoked on the HeatWave cluster.

• last_updated_timestamp

  The POSIX timestamp of the last call.