Query Pipeline
Query Engine
Trace the end-to-end journey of a TSDB query — from label matching through postings intersection, chunk pruning, stats shortcuts, decoding, and step-aligned aggregation. Build a query and watch it execute stage by stage.
① Sample Dataset
A simulated TSDB with three metrics, multiple label dimensions, and chunked storage.
The data model: A monitoring system stores series — each series is one metric stream identified by labels (key=value pairs). Each series is split into fixed-size chunks of compressed samples. A medium-sized system might have 50,000 series, each split into hundreds of chunks.
② Query Pipeline
Six stages transform a PromQL-like selector into aggregated results. Each stage lights up as the query executes.
③ Query Builder
Configure a query and hit Execute to watch the pipeline animate.
④ Label Matching & Postings Intersection
How label lookups work: For each unique label value (e.g. region=us-west-2), the TSDB maintains a postings list — a sorted array of series IDs that have that label. To find series matching multiple filters, we intersect these lists: keep only IDs that appear in all of them.