Large-Scale Performance Data Retrieval

The example described in the previous sections shows how to retrieve performance data for a single entity. When you design your application to retrieve performance data on a large scale, take the following information into consideration for more efficient processing.

• Use CSV formatted output. CSV format provides a more compact representation of the output data which can save on meta-data overhead.
• Create query specifications to reference a set of vSphere entities.
  • Using one QueryPerf method call per entity is not efficient.
  • Using a single call to QueryPerf to retrieve all of the performance data is not efficient.
  • As a general rule, specify between 10 and 50 entities in a single call to the QueryPerf method. This is a general recommendation because your system configuration may impose different constraints.
• Do not retrieve statistics more frequently than they are refreshed. For example, when you retrieve 20-second interval data, the data will not change until the next 20-second data collection event.
• Use QueryAvailablePerfMetric only when you intend to send a query for a specific counter using a specific performance interval. The method will return PerfMetricId objects that you can use for the query.

  In all other cases, create the PerfMetricId objects for the query.

  • For the counterId property, use the counter IDs from the PerformanceManager counter list (PerformanceManager.perfCounter[].key).
  • For the instance property, specify an asterisk (“*”) to retrieve instance and aggregate data or a zero-length string (““) to retrieve aggregate data only.