UsageData is defined as follows:

public class UsageData {    /* the usage type for this resource. Should match the type specified in the metadata      */    String getUsageType();    /* The tenant handle if any. Should match the handle for owning tenant associated        in the account lifecycle handler */    String getTenantHandle();    /* The user handle if any. Should match the handle for the owning user associated        in the user lifecycle handler */    String getUserHandle();    /* The resource handle, if any. Should match the handle for this tenant        associated in the subscription lifecycle handler */    String getResourceHandle();    /* Start date/time for this usage record */    Date getStartDate();    /* End Date/Time for this usage record */    Date getEndDate();    /* service specific reference id for this usage record. This should be the same        format used by the markers. */    String getReferenceId();    /* Discriminator values for *all* declared discriminators for this usage. This        includes resouce type ids, filter values and other discriminators declared for       this usage type in the metadata */    Map<String, String> getDisriminators ();    /* Raw usage. Actual usage consumed, in the unit of measure described in metadata */    Double getUsage();  }

The UsageCollector SPI defines the protocol which CPBM uses to collect usage from the underlying service on a periodic basis. It is defined as follows:

public interface UsageCollector {    /* Returns a batch of usage data collected from the cloud service. */    public UsageResultSet collect(String startMarker, Date endDate);  }

The UsageCollector is called periodically to collect data from the underlying cloud service. CPBM will schedule a job to collect this data at an operator defined frequency. In any given run CPBM will repeatedly call the UsageCollector to collect usage, until it exhausts available data. The startMarker and endDate determines the range of usage data to collect. When collect is called, the returned usage data:

• should start from after the startMarker and it must be generated on or before the passed endDate. For better performance, we recommend all implementations of the UsageCollector return usage data in small batches. The batch size should typically be in hundreds. Any subsequent call to UsageCollector.collect() should return the next batch of usageData. In case there is no usageData left, return UsageResultSet with the input startMarker as the marker and null usageData list. CPBM will continuously call UsageCollector.collect() for a given endDate, until all data is exhausted for this time range.
• In case of conflicting usageData (i.e., duplicate referenceId) CPBM will resolve the conflict by using the first record received.
• The UsageResultSet object contains List of UsageData. This list should be sorted by UsageData.endDate from oldest to newest. Also, for any two calls to UsageCollector.collect, all of UsageData collected in the later call should be newer (have later UsageData.endDate, than the data collected in the earlier call).