[BEAM-11791] Adding a microbenchmark for TestStream #14046
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
pabloem
force-pushed
the
teststream-ubench
branch
from
d0c0cf4 to
7a6b6bb
Compare
Member
Author
|
Run Python Load Tests FnApiRunner Microbenchmark |
Member
Author
Member
Author
|
Run Python PreCommit |
rohdesamuel
suggested changes
Feb 24, 2021
| def run_benchmark( | ||
| starting_point=1, num_runs=10, num_elements_step=300, verbose=True): | ||
| suite = [ | ||
| utils.LinearRegressionBenchmarkConfig( |
Contributor
There was a problem hiding this comment.
I'm confused, why is this running a linear regression? Shouldn't it just be a windowed average over the last N runs?
Member
Author
There was a problem hiding this comment.
This runs a pipeline with 1 element, then 300, then 600, and so on until 3000. We use the linear regression to measure per-element overhead, and base overhead of the runner, and report it.
Contributor
There was a problem hiding this comment.
Still feels very weird to me to be using a regression. Usually those are meant to build models/feature extraction, but this seems the benchmark only needs to report some windowed average. Anyway, that's a conversation for another time.
| 3) When the timer fires, change the key and output all the elements downstream | ||
|
|
||
| This executes the same codepaths that are run on the Fn API (and Dataflow) | ||
| workers, but is generally easier to run (locally) and more stable.. |
Contributor
There was a problem hiding this comment.
Remove extraneous period
Comment on lines
26
to
28
| 1) Put all the PCollection elements in state | ||
| 2) Set a timer for the future | ||
| 3) When the timer fires, change the key and output all the elements downstream |
Contributor
There was a problem hiding this comment.
This comment is somewhat misleading. Technically, windows are state & timers but I don't know if many people will make that connection reading this code. Can you please rewrite what this microbenchmark does in terms of windowing?
Member
Author
There was a problem hiding this comment.
Thanks. Great catch. I've rephrased this. LMK if htat looks better.
Member
Author
|
thanks Sam! |
kennknowles
added a commit
to kennknowles/beam
that referenced
this pull request
Mar 4, 2021
…gger * github/master: (123 commits) [BEAM-11899] Bump commons-pool to 2.8.1 and bump commons-dbcp to 2.8.0, Because there is a library dependency Update pillars.yaml (apache#14142) [BEAM-10632] Checkerframework nullness cleanups (apache#14107) [BEAM-11213] Instantiate SparkListenerApplicationStart in a Spark 3 compatible way Fix typos for excluding testMergingCustomWindowsWithoutCustomWindowTypes Specify the time resolution for TestStreamPayload. [BEAM-10961] Enable strict depdency checking for sdks/java/extensions/euphoria (second attempt) [BEAM-11848] Store Docker images in a variable for consistency. Splitting old Go Precommit and new ULR integration test precommit. Moving runner imports out of ptest. Add the TO_STRING capability to Java and Python [BEAM-11848] Fix Docker images list. jdbc python supported Dataflow runner (apache#13960) Adding a warning to use multi-workers on FnApiRunner Fix legend for Python Directrunner microbenchmarks [BEAM-11740] Estimate PCollection byte size [BEAM-10961] enable strict dependency checking for sdks/java/extensions/zetasketch (apache#14093) Map Dataflow JOB_STATE_CANCELLING to Beam RUNNING state [BEAM-11833] Fix reported watermark after restriction split in UnboundedSourceAsSDFRestrictionTracker [BEAM-10761] add reference to BEAM-11761 [BEAM-10961] enable strict dependency checking for flink/job-server Exclude MapState example integration tests from Dataflow runner v2 suite Remove InvalidWindows from Java SDK, instead track "already merged" bit Fix checkstyle in watermark latency benchmark Fix compile breakage in WindmillStateInternals Improve test, error on ALREADY_MERGED. [BEAM-10961] Strict dependency checking for sdks/java/io/gcp (apache#13791) Initial watermark latency benchmark Attempting improvements on DirectRunner Python dash [BEAM-10961] enable strict dependency checking for sdks/java/extensions/google-cloud-platform-core (apache#14084) Merge pull request apache#13802: [BEAM-1474]. Adding MapState and SetState support for the Dataflow runner Remove some false positives Remove nullness warning suppression [BEAM-11861] Add methods to explicitly provide coder for ParquetIO's Parse and ParseFiles (apache#14078) [BEAM-11531] Use pandas 1.2 for python>=3.7 (apache#14099) [BEAM-10961] add reference to BEAM-11761 [BEAM-10961] add explicit compile for auto_value_annotations in sdks/extensions/ml/build.gradle Attempting improvements on DirectRunner Python dash Recognize JOB_STATE_PENDING from Dataflow and map to RUNNING never run checkerframework on tests Puts more expensive BQ empty table check to the right of the 'and' condition (apache#14094) Use the windowing strategy of the input, not output, PCollection of GBK. Do not stage dataflow worker jar when use runner_v2. [BEAM-11870] Re-raise underlying exception for InvocationTargetException (apache#14098) [BEAM-11778] Create a wrapper for ZetaSQL catalog and refactor accordingly. (apache#13934) [BEAM-9378] Add ignored tests which fail in various ways when querying nested structures (apache#14077) Merge Fn API and runner v2 configurations for DataflowRunner Fix up! formatting Add validate runner test for testing custom merging windows fn without custom window types Revert "Revert "[BEAM-2914] Add portable merging window support to Python. (apache#12995)"" [BEAM-10961] fix stray reordering of lines [BEAM-10961] enable strict dependency checking for sdks/java/extensions/sorter [BEAM-10961] enable strict dependency checking for sdks/java/extensions/sketching [BEAM-10961] enable strict dependency checking for sdks/java/extensions/schemaio-expansion-service [BEAM-10961] enable strict dependency checking for sdks/java/extensions/protobuf [BEAM-10961] enable strict dependency checking for sdks/java/extensions/ml [BEAM-10961] enable strict dependency checking for sdks/java/extensions/kyro [BEAM-10961] enable strict dependency checking for sdks/java/extensions/join-library [BEAM-10961] undo line moves (originally intended for alphabeticization) [BEAM-10961] enable strict dependency checking for sdks/java/extensions/jackson [BEAM-10961] Enable strict dependency checking on sdks/java/extensions/sql (apache#13830) [BEAM-10961] enable strict dependency checking for sdks/java/extensions/euphoria [BEAM-10961] enable strict dependency checking for sdks/java/io/parquet (apache#14062) [BEAM-10961] enable strict dependency checking for sdks/java/io/thrift (apache#14066) Refactor ZetaSqlDialectSpecTest and add some passing tests. (apache#14080) [BEAM-11864] Use objects.equals instead of raw comparison [BEAM-11707] Change WindmillStateCache cache invalidation to be based upon reference invalidation instead of expensive set management. Reduce operations of shared cache by caching per-key object sets locally and flushing as groups to shared cache. Remove byte tracking which could be racy based upon background evictions in favor of just iterating for rendering the status page. This also lets us capture more stats. [BEAM-11730] Reduce context switching overhead for appliance reads by issuing reads directly from calling threads in the case that there is no reads being queued. Fix preview Show string from Dataflow service when job terminates in unrecognized state Log a warning when Dataflow returns an unrecognized state Merge pull request apache#14033 from [BEAM-11408] Integrate Python BigQuery sink with GroupIntoBatches Remove SYNCHRONIZED_PROCESSING_TIME from model proto Remove use of model SYNCHRONIZED_PROCESSING_TIME Merge redundant model feature columns in capability matrix Remove MapReduce runner from capability matrix, because it is on a branch and unreleased Remove JStorm runner from capability matrix, because it is on a branch and unreleased Remove retractions from capability matrix, because they do not exist yet Remove metadata-driven triggers from capability matrix, because they do not exist [BEAM-10937] Add Tour of Beam page (apache#13747) [BEAM-11344] Apply "Become a Committer" changes from Website Revamp (apache#14036) Merge pull request apache#14046 from [BEAM-11791] Adding a microbenchmark for TestStream Returning successful writes in FhirIO.Write.Result (apache#14034) Fixup [BEAM-10961] enable strict dependency checking for sdks/java/io/file-based-io-tests (apache#14052) [BEAM-10961] enable strict dependency checking for sdks/java/io/contextualtextio (apache#14049) [BEAM-10961] enable strict dependency checking for sdks/java/io/kinesis (apache#14058) [BEAM-10961] enable strict dependency checking for sdks/java/io/bigquery-io-perf-tests (apache#14048) [BEAM-10961] enable strict dependency checking for sdks/java/io/elasticsearch (apache#14050) [BEAM-10961] enable strict dependency checking for sdks/java/io/expansion-service (apache#14051) [BEAM-10961] enable strict dependency checking for sdks/java/io/jdbc (apache#14055) [BEAM-10961] enable strict dependency checking for sdks/java/io/jms (apache#14056) [BEAM-10961] enable strict dependency checking for sdks/java/io/kafka (apache#14057) [BEAM-10961] enable strict dependency checking for sdks/java/io/hcatalog (apache#14053) [BEAM-11859] Fixed bug in python S3 IO [BEAM-10114] Fix PerSubscriptionPartitionSdf to not rely on the presence of BundleFinalizer [BEAM-10114] Fix PerSubscriptionPartitionSdf to not rely on the presence of BundleFinalizer [BEAM-10961] fix spacing [BEAM-10961] enable strict dependency checking for sdks/java/io/xml [BEAM-10961] enable strict dependency checking for sdks/java/io/tika ...
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Adding a microbenchmark for running a simple pipeline with TestStream and a few groupings in parallel.
r: @rohdesamuel
Thank you for your contribution! Follow this checklist to help us incorporate your contribution quickly and easily:
R: @username).[BEAM-XXX] Fixes bug in ApproximateQuantiles, where you replaceBEAM-XXXwith the appropriate JIRA issue, if applicable. This will automatically link the pull request to the issue.CHANGES.mdwith noteworthy changes.See the Contributor Guide for more tips on how to make review process smoother.
Post-Commit Tests Status (on master branch)
Pre-Commit Tests Status (on master branch)
See .test-infra/jenkins/README for trigger phrase, status and link of all Jenkins jobs.
GitHub Actions Tests Status (on master branch)
See CI.md for more information about GitHub Actions CI.