Today the TSS Integration Layer expects to get the full Metric for each data point back from the database:
This has 2 downsides:
- it might be a lot of data to retrieve
- we need to reconstruct the tags in order to recreate the metric.
This should be unnecessary since we have already the Metric (part of the TimeSeriesFetchRequest).
So why won't we remove the necessity to query for the Metric?
Because we need it to get the type of the metric (counter, gauge, etc) in order to aggregate when using the Newts aggregation. Maybe we can remove that dependency and therefor drop the need to load the metric at all?