Ensure ordinal builder emit ordinal blocks #127949
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Hi @dnhatn, I've created a changelog YAML for you. |
elasticsearchmachine
added
the
Team:Analytics
|
Pinging @elastic/es-analytical-engine (Team:Analytics) |
martijnvg
approved these changes
May 9, 2025
kkrik-es
reviewed
May 9, 2025
| @@ -55,7 +59,7 @@ public SingletonOrdinalsBuilder endPositionEntry() { | |||
| } | |||
|
|
|||
| BytesRefBlock buildOrdinal() { | |||
| int valueCount = docValues.getValueCount(); | |||
| int valueCount = maxOrd - minOrd + 1; | |||
There was a problem hiding this comment.
Should this be changed? The values are not guaranteed to be dense, iiuc?
There was a problem hiding this comment.
Yes, this is the estimate. I will work on on a follow-up to resolve this completely.
kkrik-es
approved these changes
May 9, 2025
16 tasks
dnhatn
added
the
auto-backport
This was referenced May 9, 2025
dnhatn
added a commit
to dnhatn/elasticsearch
that referenced
this pull request
May 9, 2025
Currently, if a field has high cardinality, we may mistakenly disable
emitting ordinal blocks. For example, with 10,000 `tsid` values, we
never emit ordinal blocks during reads, even though we could emit blocks
for 10 `tsid` values across 1,000 positions. This bug disables
optimizations for value aggregation and block hashing.
This change tracks the minimum and maximum seen ordinals and uses them
as an estimate for the number of ordinals. However, if a page contains
`ord=1` and `ord=9999`, ordinal blocks still won't be emitted.
Allocating a bitset or an array for `value_count` could track this more
accurately but would require additional memory. I need to think about
this trade off more before opening another PR to fix this issue
completely.
This is a quick, contained fix that significantly speeds up time-series
aggregation (and other queries too).
The execution time of this query is reduced from 3.4s to 1.9s with 11M documents.
```
POST /_query
{
"profile": true,
"query": "TS metrics-hostmetricsreceiver.otel-default
| STATS cpu = avg(avg_over_time(`metrics.system.cpu.load_average.1m`)) BY host.name, BUCKET(@timestamp, 5 minute)"
}
```
```
"took": 3475,
"is_partial": false,
"documents_found": 11368089,
"values_loaded": 34248167
```
```
"took": 1965,
"is_partial": false,
"documents_found": 11368089,
"values_loaded": 34248167
```
dnhatn
added a commit
to dnhatn/elasticsearch
that referenced
this pull request
May 9, 2025
Currently, if a field has high cardinality, we may mistakenly disable
emitting ordinal blocks. For example, with 10,000 `tsid` values, we
never emit ordinal blocks during reads, even though we could emit blocks
for 10 `tsid` values across 1,000 positions. This bug disables
optimizations for value aggregation and block hashing.
This change tracks the minimum and maximum seen ordinals and uses them
as an estimate for the number of ordinals. However, if a page contains
`ord=1` and `ord=9999`, ordinal blocks still won't be emitted.
Allocating a bitset or an array for `value_count` could track this more
accurately but would require additional memory. I need to think about
this trade off more before opening another PR to fix this issue
completely.
This is a quick, contained fix that significantly speeds up time-series
aggregation (and other queries too).
The execution time of this query is reduced from 3.4s to 1.9s with 11M documents.
```
POST /_query
{
"profile": true,
"query": "TS metrics-hostmetricsreceiver.otel-default
| STATS cpu = avg(avg_over_time(`metrics.system.cpu.load_average.1m`)) BY host.name, BUCKET(@timestamp, 5 minute)"
}
```
```
"took": 3475,
"is_partial": false,
"documents_found": 11368089,
"values_loaded": 34248167
```
```
"took": 1965,
"is_partial": false,
"documents_found": 11368089,
"values_loaded": 34248167
```
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Currently, if a field has high cardinality, we may mistakenly disable emitting ordinal blocks. For example, with 10,000
tsidvalues, we never emit ordinal blocks during reads, even though we could emit blocks for 10tsidvalues across 1,000 positions. This bug disables optimizations for value aggregation and block hashing.This change tracks the minimum and maximum seen ordinals and uses them as an estimate for the number of ordinals. However, if a page contains
ord=1andord=9999, ordinal blocks still won't be emitted. Allocating a bitset or an array forvalue_countcould track this more accurately but would require additional memory. I need to think about this trade off more before opening another PR to fix this issue completely.This is a quick, contained fix that significantly speeds up time-series aggregation (and other queries too).
The execution time of this query is reduced from 3.4s to 1.9s with 11M documents.