The “Indexed” Outcome Category

Data types which are subject to dynamic sampling are represented in two different outcome data categories: "total" and "indexed". You can think of the "total" category as counting items that are stored in aggregated form, in other words, as metrics. The "indexed" category counts items that are stored individually.

If an item is dropped before metrics extraction & dynamic sampling, it will show up as an outcome in both categories (for example, "transaction" and "transaction_indexed"). If an item is dropped by dynamic sampling, it will show up as a "filtered" outcome in the "indexed" category, but it lives on in the form of metrics, for which we emit an "accepted" outcome in the "total" category.

Illustration

These charts illustrate the flow of data categories for transactions:

For a sampled transaction:

flowchart LR
  SDK --transaction--> Relay
  Relay --transaction--> DS[Dynamic Sampling]
  Relay --transaction_indexed--> DS
  DS --transaction-->Store
  DS --transaction_indexed-->Store

For a transaction filtered by dynamic sampling:

flowchart LR
  SDK --transaction--> Relay
  Relay --transaction--> DS[Dynamic Sampling]
  Relay --transaction_indexed--> DS
  DS --transaction-->Store
  DS --transaction_indexed-->Reject
  style Reject stroke:#f00

Data Types

The following data categories have a corresponding "*_indexed" category:

• transactions
• spans
• profiles