General performance improvements #598
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
Signed-off-by: Oleh Dokuka <[email protected]>
OlegDokuka
requested review from
nebhale,
smaldini,
yschimke,
rdegnan and
robertroeser
Signed-off-by: Oleh Dokuka <[email protected]>
yschimke
reviewed
Mar 17, 2019
| @@ -490,8 +493,7 @@ private void handleFrame(int streamId, FrameType type, ByteBuf frame) { | |||
| LimitableRequestPublisher sender = senders.get(streamId); | |||
| if (sender != null) { | |||
| int n = RequestNFrameFlyweight.requestN(frame); | |||
| sender.increaseRequestLimit(n); | |||
| sendProcessor.drain(); | |||
| sender.request(n >= Integer.MAX_VALUE ? Long.MAX_VALUE : n); | |||
There was a problem hiding this comment.
tangential: I'm trying to remember where we specify this rule? Integer.MAX_VALUE == Inifinite? Reactor? Reactive Streams? RSocket?
There was a problem hiding this comment.
There was a problem hiding this comment.
I guess what I'm asking is whether this rule needs to migrate into the spec
There was a problem hiding this comment.
If it is not clarify that Int. Max Value should be considered as unbounded stream, then we have to mention that
yschimke
reviewed
Mar 17, 2019
rsocket-core/src/test/java/io/rsocket/fragmentation/FragmentationDuplexConnectionTest.java
Outdated
Show resolved
Hide resolved
|
Nice perf win! |
yschimke
approved these changes
Mar 17, 2019
Signed-off-by: Oleh Dokuka <[email protected]>
Signed-off-by: Oleh Dokuka <[email protected]>
Signed-off-by: Oleh Dokuka <[email protected]>
OlegDokuka
force-pushed
the
feature/performance-improvements
branch
from
f282b76 to
12669c7
Compare
Signed-off-by: Oleh Dokuka <[email protected]>
mostroverkhov
added a commit
that referenced
this pull request
Apr 5, 2019
This reverts commit f55cf8f
mostroverkhov
added a commit
that referenced
this pull request
Apr 5, 2019
This reverts commit f55cf8f Signed-off-by: Maksym Ostroverkhov <[email protected]>
robertroeser
pushed a commit
that referenced
this pull request
Apr 7, 2019
* channel load test * Revert "General performance improvements (#598)" This reverts commit f55cf8f * use byte-count for frame positions Signed-off-by: Maksym Ostroverkhov <[email protected]> * resume token is ByteBuf Signed-off-by: Maksym Ostroverkhov <[email protected]> * Keep-alive connection, receiving remote implied position: replace Flux with simple onRemoteImpliedPosition(long) callback Signed-off-by: Maksym Ostroverkhov <[email protected]> * better performance with resume on Signed-off-by: Maksym Ostroverkhov <[email protected]> * keep alive handler: less allocations Signed-off-by: Maksym Ostroverkhov <[email protected]> * remove ResumptionState to reduce allocations Signed-off-by: Maksym Ostroverkhov <[email protected]> * remove ResumedFramesCalculator Signed-off-by: Maksym Ostroverkhov <[email protected]> * rename ResumeAwareConnection to ResumePositionsConnection Signed-off-by: Maksym Ostroverkhov <[email protected]> * ResumableDuplexConnection: use drain-queue for actions Signed-off-by: Maksym Ostroverkhov <[email protected]> * tests Signed-off-by: Maksym Ostroverkhov <[email protected]> * formatter Signed-off-by: Maksym Ostroverkhov <[email protected]> * cleanup: resume token is released in ClientRSocketSession instead of RSocketFactory KeepAliveConnection: dispose KeepAliveHandler in onClose() Signed-off-by: Maksym Ostroverkhov <[email protected]> * reduce allocations Signed-off-by: Maksym Ostroverkhov <[email protected]> * resume: fix race between send / release of frames on received implied position from keepalives replace state enum with constants remove unnecessary allocation make cleanup of store on keep-alive optional formatter Signed-off-by: Maksym Ostroverkhov <[email protected]> * remove unintended changes Signed-off-by: Maksym Ostroverkhov <[email protected]> * optimize KeepAliveConnection optimize ResumableDuplexConnection Signed-off-by: Maksym Ostroverkhov <[email protected]>
mostroverkhov
added a commit
that referenced
this pull request
May 16, 2019
This reverts commit f55cf8f
mostroverkhov
added a commit
that referenced
this pull request
May 16, 2019
This reverts commit f55cf8f Signed-off-by: Maksym Ostroverkhov <[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 PR provides a list of dramatical performance improvements:
firstelements strategy. In all of those cases, usage of atomic was redundant since Reactive-Streams spec and Reactors implementation guarantees happens-before and serialized access to the object (Subscriber / Subscription) so having an additional object with volatile access is redundant and give more overheadBaseSubscriber. Thus, this PR eliminates redundant transformations and minimize function flow. Also, dynamic invocations give its overhead, so there are upcoming changes in Reactor 3 (Macro-fusion in primitive operators reactor/reactor-core#1556) that improves performance as well.General bugfixes and improvements
Along with performance improvements, this PR introduces Jmh tests so you can run benchmarks locally and observe 10-20% perf improvements comparing to the previous implementation.
Finally, this PR provides a couple of bugfixes found during development and testing
Signed-off-by: Oleh Dokuka [email protected]