Refine algorithm for WriteAmpBasedRateLimiter #213
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
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
This reverts commit dc096f2. Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Connor1996
approved these changes
Dec 15, 2020
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: tabokie <[email protected]>
Signed-off-by: Xinye Tao <[email protected]>
yiwu-arbug
reviewed
Dec 17, 2020
yiwu-arbug
reviewed
Dec 17, 2020
Signed-off-by: tabokie <[email protected]>
tabokie
added a commit
to tabokie/rocksdb
that referenced
this pull request
May 10, 2022
* Bugfix
Only compaction triggers auto-tuner to collect necessary data for training rate limit. When compaction frequency is low, data from long period of time is fused into one sample, causing inaccurate estimation. Fix this issue by looping through missing timeslice.
Recent window size (10s) is too small, make it 30s.
* Better support for low pressure scenarios
Before this PR, flush flow is padded to 20MB/s which makes rate limit always larger than 28MB/s. After removing this restriction, we notice that it's easier to accumulate pending bytes under low pressure. Adjust the padding calculation to partially resolve this problem.
Also notice that with new formula, the minimal rate limit is still around 28MB/s.
* Control reshuffle
Remove the use of long term sampler, instead enlarge the window of short term sampler. Reduce the use of `ratio_delta` which often causes unnecessary jitters. With algorithm being simplified, we can now deduce the actual limit by prometheus expression `sum(rate(tikv_engine_compaction_flow_bytes{instance=~"$instance", db="kv", type="bytes_written"}[5m]))`
* Normal pace up
Add normal pace in addition to critical pace up to reduce pending bytes issue.
Signed-off-by: tabokie <[email protected]>
tabokie
added a commit
that referenced
this pull request
May 11, 2022
* Bugfix
Only compaction triggers auto-tuner to collect necessary data for training rate limit. When compaction frequency is low, data from long period of time is fused into one sample, causing inaccurate estimation. Fix this issue by looping through missing timeslice.
Recent window size (10s) is too small, make it 30s.
* Better support for low pressure scenarios
Before this PR, flush flow is padded to 20MB/s which makes rate limit always larger than 28MB/s. After removing this restriction, we notice that it's easier to accumulate pending bytes under low pressure. Adjust the padding calculation to partially resolve this problem.
Also notice that with new formula, the minimal rate limit is still around 28MB/s.
* Control reshuffle
Remove the use of long term sampler, instead enlarge the window of short term sampler. Reduce the use of `ratio_delta` which often causes unnecessary jitters. With algorithm being simplified, we can now deduce the actual limit by prometheus expression `sum(rate(tikv_engine_compaction_flow_bytes{instance=~"$instance", db="kv", type="bytes_written"}[5m]))`
* Normal pace up
Add normal pace in addition to critical pace up to reduce pending bytes issue.
Signed-off-by: tabokie <[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.
Bugfix
Only compaction triggers auto-tuner to collect necessary data for training rate limit. When compaction frequency is low, data from long period of time is fused into one sample, causing inaccurate estimation. Fix this issue by looping through missing timeslice.
Recent window size (10s) is too small, make it 30s.
Better support for low pressure scenarios
Before this PR, flush flow is padded to 20MB/s which makes rate limit always larger than 28MB/s. After removing this restriction, we notice that it's easier to accumulate pending bytes under low pressure. Adjust the padding calculation to partially resolve this problem.
Also notice that with new formula, the minimal rate limit is still around 28MB/s.
Control reshuffle
Remove the use of long term sampler, instead enlarge the window of short term sampler. Reduce the use of
ratio_deltawhich often causes unnecessary jitters. With algorithm being simplified, we can now deduce the actual limit by prometheus expressionsum(rate(tikv_engine_compaction_flow_bytes{instance=~"$instance", db="kv", type="bytes_written"}[5m]))