Skip to content

Commit 4b2a4c9

Browse files
authored
Merge pull request #4662 from vespa-engine/bratseth/document-threadpool-size
Document the default for search threads on cloud
2 parents 84ccaf7 + f4693b7 commit 4b2a4c9

1 file changed

Lines changed: 7 additions & 7 deletions

File tree

en/reference/applications/services/content.html

Lines changed: 7 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -752,7 +752,7 @@ <h2 id="requestthreads">requestthreads</h2>
752752
<tr>
753753
<th>search</th>
754754
<td>Optional</td>
755-
<td><strong>Vespa Cloud:</strong> number of allocated vCPUs.
755+
<td><strong>Vespa Cloud:</strong> min(vcpu*4 + persearch - 1, vcpu*persearch)<br/>
756756
<strong>Self-hosted:</strong> 64.</td>
757757
<td><p id="requestthreads-search">Total size of the match engine thread pool.
758758
Together with <code>persearch</code>, this determines the maximum number of queries
@@ -766,10 +766,10 @@ <h2 id="requestthreads">requestthreads</h2>
766766
<td>1</td>
767767
<td><p id="requestthreads-persearch">
768768
Maximum number of threads used per search.
769-
A higher value reduces query latency for queries where the
770-
<a href="../../../performance/sizing-search.html#dynamic-query-work">dynamic query work</a> is high,
771-
but also reduces the maximum number of concurrent queries
772-
(<code>search / persearch</code>).
769+
A higher value reduces the time queries spend in query evaluation, except
770+
time spent in ANN which is single-threaded.
771+
This number of threads is held for each query for the duration of the query,
772+
also when much of the time is spent on single-threaded operations.
773773
See the <a href="../../../performance/sizing-search.html">Vespa serving scaling guide</a>
774774
for an introduction of using multiple threads per search per node to reduce query latency.
775775
Number of threads per search can be adjusted down per <em>rank-profile</em>
@@ -778,8 +778,8 @@ <h2 id="requestthreads">requestthreads</h2>
778778
<tr>
779779
<th>summary</th>
780780
<td>Optional</td>
781-
<td><strong>Vespa Cloud:</strong> number of allocated vCPUs.
782-
<strong>Self-hosted:</strong> 16.</td>
781+
<td><strong>Vespa Cloud:</strong> vcpu
782+
<strong>Self-hosted:</strong> 16</td>
783783
<td><p id="requestthreads-summary">Number of summary threads.</p></td>
784784
</tr>
785785
</tbody>

0 commit comments

Comments
 (0)