db: optimize SeekGE/SeekLT for the noop case #1067
Merged
Conversation
This file contains 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
This is especially useful for sparse parts of the key space which may have large numbers of deleted versions, as happens with the locks in CockroachDB. Results from a CockroachDB benchmark are included below. The separated=true case is what mainly benefits from this. The actual new numbers are listed after the diff. name old time/op new time/op delta ScanAllIntentDeleted/separated=false/versions=200/percent-flushed=0-16 63.8µs ± 0% 50.2µs ± 0% -21.27% ScanAllIntentDeleted/separated=false/versions=200/percent-flushed=50-16 24.4µs ± 0% 24.8µs ± 0% +1.81% ScanAllIntentDeleted/separated=false/versions=200/percent-flushed=90-16 5.60µs ± 0% 4.28µs ± 0% -23.68% ScanAllIntentDeleted/separated=false/versions=200/percent-flushed=100-16 1.59µs ± 0% 1.07µs ± 0% -33.04% ScanAllIntentDeleted/separated=true/versions=200/percent-flushed=0-16 51.7ms ± 0% 0.0ms ± 0% -99.99% ScanAllIntentDeleted/separated=true/versions=200/percent-flushed=50-16 14.7ms ± 0% 0.0ms ± 0% -99.99% ScanAllIntentDeleted/separated=true/versions=200/percent-flushed=90-16 1.04ms ± 0% 0.00ms ± 0% -99.85% BenchmarkScanAllIntentDeleted/separated=false/versions=200/percent-flushed=0-16 23816 50222 ns/op 0 B/op 0 allocs/op BenchmarkScanAllIntentDeleted/separated=false/versions=200/percent-flushed=50-16 46825 24840 ns/op 0 B/op 0 allocs/op BenchmarkScanAllIntentDeleted/separated=false/versions=200/percent-flushed=90-16 304212 4277 ns/op 0 B/op 0 allocs/op BenchmarkScanAllIntentDeleted/separated=false/versions=200/percent-flushed=100-16 1342129 1066 ns/op 0 B/op 0 allocs/op BenchmarkScanAllIntentDeleted/separated=true/versions=200/percent-flushed=0-16 197700 5267 ns/op 0 B/op 0 allocs/op BenchmarkScanAllIntentDeleted/separated=true/versions=200/percent-flushed=50-16 673788 1696 ns/op 0 B/op 0 allocs/op BenchmarkScanAllIntentDeleted/separated=true/versions=200/percent-flushed=90-16 798824 1530 ns/op 0 B/op 0 allocs/op BenchmarkScanAllIntentDeleted/separated=true/versions=200/percent-flushed=100-16 1256680 1027 ns/op 0 B/op 0 allocs/op
|
This change is |
jbowens
approved these changes
Feb 19, 2021
There was a problem hiding this comment.
Nice.
Reviewable status: 0 of 1 files reviewed, all discussions resolved (waiting on @itsbilal and @petermattis)
itsbilal
approved these changes
Feb 19, 2021
There was a problem hiding this comment.
Reviewed 1 of 1 files at r1.
Reviewable status:complete! all files reviewed, all discussions resolved (waiting on @jbowens and @petermattis)
|
TFTR! |
|
@jbowens I used your suggestion, from a previous optimization PR, of replacing |
sumeerbhola
added a commit
to sumeerbhola/cockroach
that referenced
this pull request
Feb 22, 2021
560af6e1 db: fix bug in seek noop optimization and add more testing Release note: None Release justification: A small bug fix to a Pebble change that was reverted earlier this day. The Pebble change significantly improves performance with separated intents as indicated in the original Pebble PR cockroachdb/pebble#1067 and is a prerequisite for enabling separated intents (Category 2 or 4)
sumeerbhola
added a commit
to sumeerbhola/cockroach
that referenced
this pull request
Feb 26, 2021
It stresses the case where many versions were written transactionally and all were resolved. This is the benchmark mentioned in cockroachdb/pebble#1067 to justify the Pebble seek noop optimization. Release justification: Non-production code change that adds a benchmark. Release note: None
sumeerbhola
added a commit
to sumeerbhola/cockroach
that referenced
this pull request
Mar 1, 2021
It stresses the case where many versions were written transactionally and all were resolved. This is the benchmark mentioned in cockroachdb/pebble#1067 to justify the Pebble seek noop optimization. Release justification: Non-production code change that adds a benchmark. Release note: None
craig bot
pushed a commit
to cockroachdb/cockroach
that referenced
this pull request
Mar 1, 2021
61037: streamingest: add AOST to stream ingestion job r=adityamaru,dt a=pbardea To be rebased on the PR that introduces the core changefeed stream client. Release justification: low-risk (very experimental feature), high reward (AOST stream ingestion) Release note: None 61197: execinfra: fix starting of processors r=yuzefovich a=yuzefovich Almost all processors embed `ProcessorBase` and call `StartInternal` on it. That function derives a new context and starts a tracing span if the tracing is enabled on the parent ctx and the derived context is returned. Previously, in all callsites the inputs to the processors would get `Start`ed before `StartInternal` is called, but this doesn't make sense: the consumer's ctx and span should be encompassing the producer's (input's) ctx and span, so this commit switches all callsites to the following layout ``` ctx = p.StartInternal(ctx) input.Start(ctx) // if there are any inputs ``` Release justification: bug fix. Release note: None 61198: storage: add a scan benchmark with resolved intents r=sumeerbhola a=sumeerbhola It stresses the case where many versions were written transactionally and all were resolved. This is the benchmark mentioned in cockroachdb/pebble#1067 to justify the Pebble seek noop optimization. Release justification: Non-production code change that adds a benchmark. Release note: None Co-authored-by: Paul Bardea <[email protected]> Co-authored-by: Yahor Yuzefovich <[email protected]> Co-authored-by: sumeerbhola <[email protected]>
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.
This is especially useful for sparse parts of the
key space which may have large numbers of deleted
versions, as happens with the locks in CockroachDB.
Results from a CockroachDB benchmark are included
below. The separated=true case is what mainly benefits
from this. The actual new numbers are listed after
the diff.