Query statistics

Every QueryResult carries a statistics block taken from Athena's GetQueryExecution. It is the cheapest way to see what a query cost, how long it waited, and whether it was served from cache.

const result = await athena.query<Row>(sql, { includeRuntimeStats: true });

result.statistics.totalExecutionTimeInMillis;   // 8128
result.statistics.resultReused;                 // true
result.statistics.runtime?.outputRows;          // 99

The type

Included from src/runtime/types.ts:

/**
 * What Athena reports about an execution. Every {@link QueryResult} carries
 * one. `dataScannedInBytes` is what drives the bill.
 */
export interface QueryStatistics {
  /** Engine execution time, in ms. */
  engineExecutionTimeInMillis: number;
  /** Wall time Athena took, in ms. */
  totalExecutionTimeInMillis: number;
  /** Time spent waiting in the queue, in ms. */
  queryQueueTimeInMillis: number;
  /** Planning and partition retrieval, in ms. */
  queryPlanningTimeInMillis: number;
  /** Preprocessing before the engine ran, in ms. */
  servicePreProcessingTimeInMillis: number;
  /** Result publication, in ms. */
  serviceProcessingTimeInMillis: number;
  /** Bytes scanned after partition pruning and projection. This is the cost. */
  dataScannedInBytes: number;
  /** Only present for capacity-reservation workgroups. */
  dpuCount?: number;
  /** True when Athena served the result from its result cache. */
  resultReused?: boolean;
  /** Populated only when `query()` is called with `{ includeRuntimeStats: true }`. */
  runtime?: QueryRuntimeRows;
}

Field	Meaning
engineExecutionTimeInMillis	engine execution time
totalExecutionTimeInMillis	wall time Athena took
queryQueueTimeInMillis	time spent waiting in the queue
queryPlanningTimeInMillis	planning and partition retrieval
servicePreProcessingTimeInMillis	preprocessing before the engine ran
serviceProcessingTimeInMillis	result publication
dataScannedInBytes	what drives the bill
dpuCount	capacity-reservation workgroups only
resultReused	true when Athena served the result from its cache
runtime	input/output row and byte counts, opt-in

Runtime statistics are opt-in

runtime stays undefined unless you pass { includeRuntimeStats: true }, because populating it costs an extra GetQueryRuntimeStatistics API call. Ask for it when you are measuring; leave it off in hot paths.