-
Notifications
You must be signed in to change notification settings - Fork 336
Update client-side stats to use light weight Hashtable #11382
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Open
dougqh
wants to merge
78
commits into
dougqh/conflating-metrics-background-work
Choose a base branch
from
dougqh/optimize-metric-key
base: dougqh/conflating-metrics-background-work
Could not load branches
Branch not found: {{ refName }}
Loading
Could not load tags
Nothing to show
Loading
Are you sure you want to change the base?
Some commits from the old base branch may be removed from the timeline,
and old review comments may become outdated.
Open
Changes from all commits
Commits
Show all changes
78 commits
Select commit
Hold shift + click to select a range
0294058
Add Hashtable and LongHashingUtils to datadog.trace.util
dougqh f751ab4
Add AggregateTable + AggregateEntry backed by Hashtable
dougqh f1b030a
Swap LRUCache for AggregateTable in Aggregator + route disable() clear
dougqh 3738c85
Eliminate MetricKey: inline its fields onto AggregateEntry
dougqh 46a9055
Add unit tests for Hashtable and LongHashingUtils
dougqh 031dc89
Apply spotless formatting to Hashtable and LongHashingUtils
dougqh f9e63b9
Add JMH benchmarks for Hashtable.D1 and D2
dougqh a534e4f
Add benchmark results to HashtableBenchmark header
dougqh ba66a36
Address review feedback on Hashtable
dougqh 3108941
Fix dropped argument in HashingUtils 5-arg Object hash
dougqh 1415f12
Address review feedback on Hashtable
dougqh b7cee2f
Drop reflection in iterator tests via package-private D1.buckets
dougqh 553ffb8
Fold AggregateMetric into AggregateEntry
dougqh 07ed635
Avoid capturing lambda in Aggregator.report
dougqh df58ad7
Add context-passing forEach to Hashtable.D1 and D2
dougqh 9c6e95c
Move forEach loop body to Support helper
dougqh 7ab47f7
Merge branch 'dougqh/util-hashtable' into dougqh/optimize-metric-key
dougqh 590ab4a
Delegate AggregateTable.forEach to Support.forEach
dougqh 447ea33
Move bucket-head cast to Support.bucket helper
dougqh 46f147a
Merge branch 'dougqh/util-hashtable' into dougqh/optimize-metric-key
dougqh dd5e13f
Use Support.bucket and type chain walks as AggregateEntry
dougqh df7f98f
Drop d1_/d2_ prefix from per-table benchmark methods
dougqh e6ecc16
Add Hashtable.Support helpers: MAX_RATIO, insertHeadEntry, MutatingTa…
dougqh a409d5d
Merge branch 'dougqh/util-hashtable' into dougqh/optimize-metric-key
dougqh 96b40b8
Simplify AggregateTable via new Hashtable.Support helpers
dougqh 55ca204
Swap MAX_RATIO numerator/denominator pair for a single float + scaled…
dougqh 7b9eccd
Merge branch 'dougqh/util-hashtable' into dougqh/optimize-metric-key
dougqh 192de0c
Address second-round review on AggregateTable / Aggregator
dougqh 4bac439
Tighten Hashtable docs + rename MAX_CAPACITY to MAX_BUCKETS
dougqh de289a0
Dedupe chain-head splice in D1/D2 via keyHash insertHeadEntry overload
dougqh 2dd65ed
Tighten Entry.next encapsulation; doc hasNext; add D1/D2 getOrCreate
dougqh cc8d98a
Merge branch 'dougqh/util-hashtable' into dougqh/optimize-metric-key
dougqh 6a9063b
Use keyHash insertHeadEntry overload in AggregateTable.findOrInsert
dougqh dbb1702
Replace // nullable comments with @Nullable annotations on AggregateE…
dougqh 545e74c
Drop redundant load-factor comment from AggregateTable ctor
dougqh 9983a59
Import java.util.Objects in AggregateEntry instead of fully qualifying
dougqh d2e4477
Document evictOneStale cost and disable() best-effort offer
dougqh 66ec7f6
Hashtable: add missing braces and detach removed/replaced entries
dougqh be823ab
Merge branch 'dougqh/util-hashtable' into dougqh/optimize-metric-key
dougqh 10956b2
Add Hashtable and LongHashingUtils to datadog.trace.util
dougqh 035dc09
Add unit tests for Hashtable and LongHashingUtils
dougqh 7728b60
Apply spotless formatting to Hashtable and LongHashingUtils
dougqh 8cd2d86
Add JMH benchmarks for Hashtable.D1 and D2
dougqh c689ef9
Add benchmark results to HashtableBenchmark header
dougqh 75790eb
Address review feedback on Hashtable
dougqh 6056ff7
Fix dropped argument in HashingUtils 5-arg Object hash
dougqh da55021
Address review feedback on Hashtable
dougqh 8b8b088
Drop reflection in iterator tests via package-private D1.buckets
dougqh 0fde7cd
Add context-passing forEach to Hashtable.D1 and D2
dougqh 6d6c2e0
Move forEach loop body to Support helper
dougqh 268de2b
Move bucket-head cast to Support.bucket helper
dougqh 93813b9
Drop d1_/d2_ prefix from per-table benchmark methods
dougqh 11a58bf
Add Hashtable.Support helpers: MAX_RATIO, insertHeadEntry, MutatingTa…
dougqh 8f1828d
Swap MAX_RATIO numerator/denominator pair for a single float + scaled…
dougqh c0d3e26
Tighten Hashtable docs + rename MAX_CAPACITY to MAX_BUCKETS
dougqh a0978ba
Dedupe chain-head splice in D1/D2 via keyHash insertHeadEntry overload
dougqh e604a8f
Tighten Entry.next encapsulation; doc hasNext; add D1/D2 getOrCreate
dougqh e2642cd
Hashtable: add missing braces and detach removed/replaced entries
dougqh 585ca56
Rename LongHashingUtils.hashCodeX(Object) to hash(Object) for API con…
dougqh aab5b3b
Merge remote-tracking branch 'origin/master' into dougqh/optimize-met…
dougqh 88f169f
Merge branch 'dougqh/util-hashtable' into dougqh/optimize-metric-key
dougqh c2ff012
Merge branch 'dougqh/conflating-metrics-background-work' into dougqh/…
dougqh 1d1ba4d
Merge branch 'dougqh/conflating-metrics-background-work' into dougqh/…
dougqh 3d4f380
Merge branch 'dougqh/conflating-metrics-background-work' into dougqh/…
dougqh 9391c48
Use writer.finishBucket() count in bootstrap test for cascade compati…
dougqh ba0d390
Merge branch 'dougqh/conflating-metrics-background-work' into dougqh/…
dougqh f71384e
Merge branch 'dougqh/conflating-metrics-background-work' into dougqh/…
dougqh 471f8f7
Merge branch 'dougqh/conflating-metrics-background-work' into dougqh/…
dougqh 7a074e0
Merge branch 'dougqh/conflating-metrics-background-work' into dougqh/…
dougqh 3fa7ac7
Merge branch 'dougqh/conflating-metrics-background-work' into dougqh/…
dougqh df3b31d
Adapt reconcileSwapsSchemaWhenTagSetChanges to AggregateEntry shape
dougqh 8ec956d
Merge branch 'dougqh/conflating-metrics-background-work' into dougqh/…
dougqh a606692
Fix MetricsIntegrationTest entry recording call site
dougqh 913e7d7
Make ConflatingMetricAggregatorTest counter checks actually verify
dougqh 2dcea9a
Drop dead recordDurations(int, AtomicLongArray) batch API
dougqh 50b06e5
Warn about colon split in AggregateEntry.of test factory
dougqh b0f21bf
Add coverage for disable() -> ClearSignal threading path
dougqh 200028d
Merge branch 'dougqh/conflating-metrics-background-work' into dougqh/…
dougqh File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
492 changes: 492 additions & 0 deletions
492
dd-trace-core/src/main/java/datadog/trace/common/metrics/AggregateEntry.java
Large diffs are not rendered by default.
Oops, something went wrong.
103 changes: 0 additions & 103 deletions
103
dd-trace-core/src/main/java/datadog/trace/common/metrics/AggregateMetric.java
This file was deleted.
Oops, something went wrong.
117 changes: 117 additions & 0 deletions
117
dd-trace-core/src/main/java/datadog/trace/common/metrics/AggregateTable.java
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,117 @@ | ||
| package datadog.trace.common.metrics; | ||
|
|
||
| import datadog.trace.util.Hashtable; | ||
| import datadog.trace.util.Hashtable.MutatingTableIterator; | ||
| import datadog.trace.util.Hashtable.Support; | ||
| import java.util.function.BiConsumer; | ||
| import java.util.function.Consumer; | ||
|
|
||
| /** | ||
| * Consumer-side {@link AggregateEntry} store, keyed on the raw fields of a {@link SpanSnapshot}. | ||
| * | ||
| * <p>Replaces the prior {@code LRUCache<MetricKey, AggregateMetric>}. The win is on the | ||
| * steady-state hit path: a snapshot lookup is a 64-bit hash compute + bucket walk + field-wise | ||
| * {@code matches}, with no per-snapshot {@link AggregateEntry} allocation and no UTF8 cache | ||
| * lookups. The UTF8-encoded forms (formerly held on {@code MetricKey}) and the mutable counters | ||
| * (formerly held on {@code AggregateMetric}) both live on the {@link AggregateEntry} now, built | ||
| * once per unique key at insert time. | ||
| * | ||
| * <p><b>Not thread-safe.</b> The aggregator thread is the sole writer; {@link #clear()} must be | ||
| * routed through the inbox rather than called from arbitrary threads. | ||
| */ | ||
| final class AggregateTable { | ||
|
|
||
| private final Hashtable.Entry[] buckets; | ||
| private final int maxAggregates; | ||
| private int size; | ||
|
|
||
| AggregateTable(int maxAggregates) { | ||
| this.buckets = Support.create(maxAggregates, Support.MAX_RATIO); | ||
| this.maxAggregates = maxAggregates; | ||
| } | ||
|
|
||
| int size() { | ||
| return size; | ||
| } | ||
|
|
||
| boolean isEmpty() { | ||
| return size == 0; | ||
| } | ||
|
|
||
| /** | ||
| * Returns the {@link AggregateEntry} to update for {@code snapshot}, lazily creating one on miss. | ||
| * Returns {@code null} when the table is at capacity and no stale entry can be evicted -- the | ||
| * caller should drop the data point in that case. | ||
| */ | ||
| AggregateEntry findOrInsert(SpanSnapshot snapshot) { | ||
| long keyHash = AggregateEntry.hashOf(snapshot); | ||
| for (AggregateEntry candidate = Support.bucket(buckets, keyHash); | ||
| candidate != null; | ||
| candidate = candidate.next()) { | ||
| if (candidate.matches(keyHash, snapshot)) { | ||
| return candidate; | ||
| } | ||
| } | ||
| if (size >= maxAggregates && !evictOneStale()) { | ||
| return null; | ||
| } | ||
| AggregateEntry entry = AggregateEntry.forSnapshot(snapshot); | ||
| Support.insertHeadEntry(buckets, keyHash, entry); | ||
| size++; | ||
| return entry; | ||
| } | ||
|
|
||
| /** | ||
| * Unlinks the first entry whose {@code getHitCount() == 0}. | ||
| * | ||
| * <p>O(N) per call -- scans buckets in array order from the start every time. That's a regression | ||
| * from the prior {@code LRUCache}'s O(1) LRU eviction, but the semantic change is deliberate: at | ||
| * cap with all entries live, we drop the new key (and report it via {@code | ||
| * onStatsAggregateDropped}) rather than evicting an established key. The expectation is that the | ||
| * cap is sized to the steady-state working set, so eviction is rare; if a future workload runs | ||
| * persistently at cap, this is the place to consider caching a cursor across calls so the scan | ||
| * resumes where it left off. | ||
| */ | ||
| private boolean evictOneStale() { | ||
| for (MutatingTableIterator<AggregateEntry> iter = Support.mutatingTableIterator(buckets); | ||
| iter.hasNext(); ) { | ||
| AggregateEntry e = iter.next(); | ||
| if (e.getHitCount() == 0) { | ||
| iter.remove(); | ||
| size--; | ||
| return true; | ||
| } | ||
| } | ||
| return false; | ||
| } | ||
|
|
||
| void forEach(Consumer<AggregateEntry> consumer) { | ||
| Support.forEach(buckets, consumer); | ||
| } | ||
|
|
||
| /** | ||
| * Context-passing forEach. Useful for callers that want to avoid a capturing-lambda allocation on | ||
| * each invocation -- pass a non-capturing {@link BiConsumer} (typically a {@code static final}) | ||
| * plus whatever side-band state it needs as {@code context}. | ||
| */ | ||
| <T> void forEach(T context, BiConsumer<T, AggregateEntry> consumer) { | ||
| Support.forEach(buckets, context, consumer); | ||
| } | ||
|
|
||
| /** Removes entries whose {@code getHitCount() == 0}. */ | ||
| void expungeStaleAggregates() { | ||
| for (MutatingTableIterator<AggregateEntry> iter = Support.mutatingTableIterator(buckets); | ||
| iter.hasNext(); ) { | ||
| AggregateEntry e = iter.next(); | ||
| if (e.getHitCount() == 0) { | ||
| iter.remove(); | ||
| size--; | ||
| } | ||
| } | ||
| } | ||
|
|
||
| void clear() { | ||
| Support.clear(buckets); | ||
| size = 0; | ||
| } | ||
| } | ||
Oops, something went wrong.
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.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What highlighted in the comment is true. However if I'm not wrong when the aggregator start dropping (meaning that capacity is full and badly dimensioned) this will translate in a perf kill. I'm wondering if it's worth adding a field for the last scanned bucket to have the cost amortized right now