Display Query Statistics Using a Simple Search Request

This page describes how to get query statistics from a simple search request using the Search API.

This page discusses:

Display Query Statistics
Attribute Description

Display Query Statistics

You can submit a simple search request (word) to: http://<HOSTNAME>:<BASEPORT>/search-api/search?q=<SEARCH_REQUEST>.

In the <stats> node, at the end of the page, you get the result of the query and information on query processing time and slices.

Example:

<Stats status="ok" queueTime="0" queryProcessingElapsedTime="5151" queryProcessingCPUTime="3302"
      queryExecElapsedTime="3139" queryExecIndexCPUTime="768" queryExecSearcherCPUTime="1951"
      synthesisAndFullHitsElapsedTime="177566" synthesisIndexCPUTime="836" synthesisSearcherCPUTime="59"
      fullHitsIndexCPUTime="3818" fullHitsSearcherCPUTime="0" totalProcessingTime="180812" />
 - <slicesInfo>
  <SliceInfo instance="i0" slice="0" internalGeneration="117" lastCommit="1381151879238"
  lastJobId="1381151878907" />
   </slicesInfo>

Attribute Description

All attributes are expressed in microseconds.


Stats attribute	Description
`queueTime`	When the maximum number of queries processing in parallel is reached, next queries are queued until the first ones finish. This parameter gives the time the query has waited for a free slot.
`queryProcessingElapsedTime` `queryProcessingCPUTime`	Processing time is the time spent in the search server, between receiving the query from the client, and sending it to the indexes. This includes parsing, and eventually expanding the query according to the search query processors. Elapsed time and CPU time must be similar, because processing is single-threaded and mostly CPU-bound. Expansions however can use the dictionary and wait for disk IO.
`queryExecElapsedTime` `queryExecIndexCPUTime` `queryExecSearcherCPUTime`	This is the time spent collecting matching document IDs in the indexes. The Index CPU Time is the CPU spent in the indexes (all slices in parallel), and searcher CPU Time is the time spent in the search server (much lower as it only has to sort the hits).
`synthesisAndFullHitsElapsedTime`	This is the time spent making the synthesis from the top hits (either all, or top 10,000 usually). This makes lots of requests to ram-based structures in the indexes containing the facets seach document matches. The full hits consist only of retrieving the metas from the attribute groups for the top 10 results.
`synthesisIndexCPUTime` `synthesisSearcherCPUTime`	CPU Time spent in searcher and indexes for the synthesis. Indexes spend a lot of time processing ram-based structures to find the facets for each document. The searcher just stores and updates the totals, and then sorts them to find the top facets.
`fullHitsIndexCPUTime` `fullHitsSearcherCPUTime`	This is the CPU time spent to retrieve the metas for the top 10 documents from the attribute groups. CPU time is usually very low because these accesses are easy, or maybe you had no results?
`totalProcessingTime`	This is the total elapsed time to process the query and return the results, which match the sum of processing, `queryExec` and `synthesisAndFullHits` elapsed times. Do not add CPU times, as they include many CPUs in parallel so they can obviously be higher than the elapsed time.