We currently have mechanisms to capture details such as whether a statement was executed with a full table or index scan in its plan, and we're also discussing surfacing information about index joins. An equally valuable piece of information that we currently lack is whether a query involves predicates that do not have an associated index.
Predicates without an index require full table scans or large portion scans, which can significantly degrade performance. Especially in large-scale deployments, workloads can change and predicates can shift, which may lead to an increased number of queries involving non-indexed predicates.
Providing observability into this would offer the following benefits:
Help users identify queries that might benefit from additional indexes. While adding indexes has its trade-offs, having this information can enable users to make more informed decisions.
Provide internal telemetry to help measure plan efficiency.
We currently have mechanisms to capture details such as whether a statement was executed with a full table or index scan in its plan, and we're also discussing surfacing information about index joins. An equally valuable piece of information that we currently lack is whether a query involves predicates that do not have an associated index.
Predicates without an index require full table scans or large portion scans, which can significantly degrade performance. Especially in large-scale deployments, workloads can change and predicates can shift, which may lead to an increased number of queries involving non-indexed predicates.
Providing observability into this would offer the following benefits:
Jira issue: CRDB-30080