Refactor serializer sketches support

[vickenty]

What does this PR do?

Refactor serializer to more easily support new Otel flavors of distribution-like data.

Add DistributionWriter interface to allow time-series types to control how they should be serialized. It is implemented by the serializer and used by concrete time series types.

Add Distribution interface to b e used by the serializer to pass control to the time-series value to use DistributionWriter and provide data to the serializer.

Extract metric metadata from SketchSeries into DistributionMetadata to provide unified set of metadata that is required for all distribution metrics.

Motivation

Make it easier to add new distribution types, and serialize them efficiently.

Describe how you validated your changes

Additional Notes

rss-base.txt is 69263b1, main with cherry-picked benchmark.
rss-var1.txt is c32f994, returning a struct from GetDDSketchPoint.
rss-var2.txt is 2d2caae, returning plain tuple of primitives.

goos: linux
goarch: amd64
pkg: github.com/DataDog/datadog-agent/pkg/serializer/internal/metrics
cpu: Intel(R) Xeon(R) Platinum 8275CL CPU @ 3.00GHz
                       │ rss-base.txt │            rss-var1.txt            │         rss-var2.txt          │
                       │    sec/op    │   sec/op     vs base               │   sec/op     vs base          │
SketchSerialization-96    13.92m ± 0%   14.18m ± 0%  +1.89% (p=0.000 n=20)   13.92m ± 0%  ~ (p=0.495 n=20)

                       │ rss-base.txt │            rss-var1.txt             │            rss-var2.txt             │
                       │     B/op     │     B/op      vs base               │     B/op      vs base               │
SketchSerialization-96   6.694Mi ± 0%   6.694Mi ± 0%  +0.00% (p=0.000 n=20)   6.694Mi ± 0%  +0.00% (p=0.000 n=20)

                       │ rss-base.txt │          rss-var1.txt           │          rss-var2.txt           │
                       │  allocs/op   │  allocs/op   vs base            │  allocs/op   vs base            │
SketchSerialization-96    80.06k ± 0%   80.06k ± 0%  ~ (p=1.000 n=20) ¹   80.06k ± 0%  ~ (p=1.000 n=20) ¹

goos: linux
goarch: arm64
pkg: github.com/DataDog/datadog-agent/pkg/serializer/internal/metrics
                       │ rss-base.txt │            rss-var1.txt            │            rss-var2.txt            │
                       │    sec/op    │   sec/op     vs base               │   sec/op     vs base               │
SketchSerialization-64    19.86m ± 0%   19.68m ± 0%  -0.90% (p=0.000 n=20)   19.27m ± 0%  -2.94% (p=0.000 n=20)

                       │ rss-base.txt │            rss-var1.txt             │            rss-var2.txt             │
                       │     B/op     │     B/op      vs base               │     B/op      vs base               │
SketchSerialization-64   6.694Mi ± 0%   6.694Mi ± 0%  +0.00% (p=0.000 n=20)   6.694Mi ± 0%  +0.00% (p=0.000 n=20)

                       │ rss-base.txt │          rss-var1.txt           │          rss-var2.txt           │
                       │  allocs/op   │  allocs/op   vs base            │  allocs/op   vs base            │
SketchSerialization-64    80.06k ± 0%   80.06k ± 0%  ~ (p=1.000 n=20) ¹   80.06k ± 0%  ~ (p=1.000 n=20) ¹
¹ all samples are equal

[datadog-datadog-prod-us1-2]

🎯 Code Coverage (details)
• Patch Coverage: 66.42%
• Overall Coverage: 50.94% (-0.02%)

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: f823ad1 | Docs | Datadog PR Page | Give us feedback!}

[dd-octo-sts]

Files inventory check summary

File checks results against ancestor 140d4c75:

Results for datadog-agent_7.82.0~devel.git.369.f823ad1.pipeline.121273448-1_amd64.deb:

No change detected

[dd-octo-sts]

Static quality checks

✅ Please find below the results from static quality gates
Comparison made with ancestor 140d4c7
📊 Static Quality Gates Dashboard
🔗 SQG Job

Successful checks

Info

	Quality gate	Change	Size (prev → curr → max)
✅	agent_heroku_amd64	+4.06 KiB (0.00% increase, -0.06% of buffer)	308.669 → 308.673 → 315.230
✅	docker_agent_amd64	+3.28 KiB (0.00% increase, -0.04% of buffer)	806.578 → 806.581 → 813.790
✅	docker_agent_jmx_amd64	+3.28 KiB (0.00% increase, -0.05% of buffer)	997.475 → 997.478 → 1004.550
✅	docker_dogstatsd_amd64	+4.06 KiB (0.01% increase, -0.45% of buffer)	39.020 → 39.024 → 39.910
✅	docker_host_profiler_amd64	+3.46 KiB (0.00% increase, -0.03% of buffer)	305.031 → 305.034 → 317.640
✅	docker_host_profiler_arm64	+3.42 KiB (0.00% increase, -0.03% of buffer)	316.116 → 316.120 → 328.900
✅	dogstatsd_deb_amd64	+4.06 KiB (0.01% increase, -0.28% of buffer)	29.755 → 29.759 → 31.150
✅	dogstatsd_deb_arm64	+4.06 KiB (0.01% increase, -0.23% of buffer)	27.810 → 27.814 → 29.530
✅	dogstatsd_rpm_amd64	+4.06 KiB (0.01% increase, -0.28% of buffer)	29.755 → 29.759 → 31.150
✅	dogstatsd_suse_amd64	+4.06 KiB (0.01% increase, -0.28% of buffer)	29.755 → 29.759 → 31.150
✅	iot_agent_deb_amd64	+4.06 KiB (0.01% increase, -0.57% of buffer)	45.688 → 45.691 → 46.380
✅	iot_agent_deb_arm64	+4.06 KiB (0.01% increase, -0.30% of buffer)	42.402 → 42.406 → 43.720
✅	iot_agent_rpm_amd64	+4.06 KiB (0.01% increase, -0.57% of buffer)	45.688 → 45.692 → 46.380
✅	iot_agent_suse_amd64	+4.06 KiB (0.01% increase, -0.57% of buffer)	45.687 → 45.691 → 46.380

18 successful checks with minimal change (< 2 KiB)

	Quality gate	Current Size
✅	agent_deb_amd64	747.762 MiB
✅	agent_deb_amd64_fips	703.689 MiB
✅	agent_rpm_amd64	747.746 MiB
✅	agent_rpm_amd64_fips	703.673 MiB
✅	agent_rpm_arm64	723.277 MiB
✅	agent_rpm_arm64_fips	682.830 MiB
✅	agent_suse_amd64	747.746 MiB
✅	agent_suse_amd64_fips	703.673 MiB
✅	agent_suse_arm64	723.277 MiB
✅	agent_suse_arm64_fips	682.830 MiB
✅	docker_agent_arm64	807.131 MiB
✅	docker_agent_jmx_arm64	986.681 MiB
✅	docker_cluster_agent_amd64	208.088 MiB
✅	docker_cluster_agent_arm64	221.315 MiB
✅	docker_cws_instrumentation_amd64	7.447 MiB
✅	docker_cws_instrumentation_arm64	6.877 MiB
✅	docker_dogstatsd_arm64	37.180 MiB
✅	iot_agent_deb_armhf	43.191 MiB

[cit-pr-commenter-54b7da]

Regression Detector

Regression Detector Results

Metrics dashboard
Target profiles
Run ID: 002c74ac-a219-4984-9a40-c6c905efac65

Baseline: 140d4c7
Comparison: f823ad1
Diff

Optimization Goals: ✅ No significant changes detected

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
➖	quality_gate_logs	% cpu utilization	+2.29	[+1.21, +3.37]	1	Logs bounds checks dashboard
➖	quality_gate_security_idle	memory utilization	+0.20	[+0.14, +0.26]	1	Logs bounds checks dashboard
➖	quality_gate_idle_all_features	memory utilization	+0.11	[+0.07, +0.15]	1	Logs bounds checks dashboard
➖	quality_gate_security_no_fs_load	memory utilization	-0.36	[-0.46, -0.26]	1	Logs bounds checks dashboard
➖	quality_gate_security_mean_fs_load	memory utilization	-0.42	[-0.46, -0.38]	1	Logs bounds checks dashboard
➖	quality_gate_idle	memory utilization	-0.49	[-0.54, -0.44]	1	Logs bounds checks dashboard
➖	quality_gate_metrics_logs	memory utilization	-1.80	[-2.05, -1.55]	1	Logs bounds checks dashboard

Bounds Checks: ✅ Passed

perf	experiment	bounds_check_name	replicates_passed	observed_value	links
✅	quality_gate_idle	intake_connections	10/10	3 ≤ 4	bounds checks dashboard
✅	quality_gate_idle	memory_usage	10/10	146.89MiB ≤ 154MiB	bounds checks dashboard
✅	quality_gate_idle	total_bytes_received	10/10	576.25KiB ≤ 819.20KiB	bounds checks dashboard
✅	quality_gate_idle_all_features	intake_connections	10/10	3 ≤ 4	bounds checks dashboard
✅	quality_gate_idle_all_features	memory_usage	10/10	480.74MiB ≤ 495MiB	bounds checks dashboard
✅	quality_gate_idle_all_features	total_bytes_received	10/10	0.89MiB ≤ 1.25MiB	bounds checks dashboard
✅	quality_gate_logs	intake_connections	10/10	4 ≤ 6	bounds checks dashboard
✅	quality_gate_logs	memory_usage	10/10	181.95MiB ≤ 195MiB	bounds checks dashboard
✅	quality_gate_logs	missed_bytes	10/10	0B = 0B	bounds checks dashboard
✅	quality_gate_logs	total_bytes_received	10/10	264.20MiB ≤ 292MiB	bounds checks dashboard
✅	quality_gate_metrics_logs	cpu_usage	10/10	341.10 ≤ 2000	bounds checks dashboard
✅	quality_gate_metrics_logs	intake_connections	10/10	4 ≤ 6	bounds checks dashboard
✅	quality_gate_metrics_logs	memory_usage	10/10	390.84MiB ≤ 430MiB	bounds checks dashboard
✅	quality_gate_metrics_logs	missed_bytes	10/10	0B = 0B	bounds checks dashboard
✅	quality_gate_metrics_logs	total_bytes_received	10/10	0.86GiB ≤ 1.04GiB	bounds checks dashboard
✅	quality_gate_security_idle	cpu_usage	10/10	30.38 ≤ 40	bounds checks dashboard
✅	quality_gate_security_idle	memory_usage	10/10	296.26MiB ≤ 330MiB	bounds checks dashboard
✅	quality_gate_security_mean_fs_load	cpu_usage	10/10	61.63 ≤ 80	bounds checks dashboard
✅	quality_gate_security_mean_fs_load	memory_usage	10/10	271.03MiB ≤ 310MiB	bounds checks dashboard
✅	quality_gate_security_no_fs_load	cpu_usage	10/10	21.86 ≤ 40	bounds checks dashboard
✅	quality_gate_security_no_fs_load	memory_usage	10/10	279.28MiB ≤ 320MiB	bounds checks dashboard

Explanation

Confidence level: 90.00%
Effect size tolerance: |Δ mean %| ≥ 5.00%

Performance changes are noted in the perf column of each table:

• ✅ = significantly better comparison variant performance
• ❌ = significantly worse comparison variant performance
• ➖ = no significant change in performance

A regression test is an A/B test of target performance in a repeatable rig, where "performance" is measured as "comparison variant minus baseline variant" for an optimization goal (e.g., ingress throughput). Due to intrinsic variability in measuring that goal, we can only estimate its mean value for each experiment; we report uncertainty in that value as a 90.00% confidence interval denoted "Δ mean % CI".

For each experiment, we decide whether a change in performance is a "regression" -- a change worth investigating further -- if all of the following criteria are true:

1. Its estimated |Δ mean %| ≥ 5.00%, indicating the change is big enough to merit a closer look.

2. Its 90.00% confidence interval "Δ mean % CI" does not contain zero, indicating that if our statistical model is accurate, there is at least a 90.00% chance there is a difference in performance between baseline and comparison variants.

3. Its configuration does not mark it "erratic".

Replicate Execution Details

We run multiple replicates for each experiment/variant. However, we allow replicates to be automatically retried if there are any failures, up to 8 times, at which point the replicate is marked dead and we are unable to run analysis for the entire experiment. We call each of these attempts at running replicates a replicate execution. This section lists all replicate executions that failed due to the target crashing or being oom killed.

Note: In the below tables we bucket failures by experiment, variant, and failure type. For each of these buckets we list out the replicate indexes that failed with an annotation signifying how many times said replicate failed with the given failure mode. In the below example the baseline variant of the experiment named experiment_with_failures had two replicates that failed by oom kills. Replicate 0, which failed 8 executions, and replicate 1 which failed 6 executions, all with the same failure mode.

Experiment	Variant	Replicates	Failure	Logs	Debug Dashboard
experiment_with_failures	baseline	0 (x8) 1 (x6)	Oom killed		Debug Dashboard

The debug dashboard links will take you to a debugging dashboard specifically designed to investigate replicate execution failures.

❌ Retried Profiling Replicate Execution Failures (ddprof)

Note: Profiling replicas may still be executing. See the debug dashboard for up to date status.

Experiment	Variant	Replicates	Failure	Debug Dashboard
quality_gate_idle	baseline	10	Oom killed	Debug Dashboard
quality_gate_idle	comparison	10	Oom killed	Debug Dashboard
quality_gate_idle_all_features	baseline	10	Oom killed	Debug Dashboard
quality_gate_idle_all_features	comparison	10	Oom killed	Debug Dashboard
quality_gate_logs	comparison	10	Oom killed	Debug Dashboard
quality_gate_metrics_logs	baseline	10	Oom killed	Debug Dashboard
quality_gate_metrics_logs	comparison	10	Oom killed	Debug Dashboard
quality_gate_security_idle	baseline	10	Oom killed	Debug Dashboard
quality_gate_security_mean_fs_load	baseline	10	Oom killed	Debug Dashboard
quality_gate_security_mean_fs_load	comparison	10	Oom killed	Debug Dashboard
quality_gate_security_no_fs_load	baseline	10	Oom killed	Debug Dashboard
quality_gate_security_no_fs_load	comparison	10	Oom killed	Debug Dashboard

CI Pass/Fail Decision

✅ Passed. All Quality Gates passed.

• quality_gate_idle, bounds check total_bytes_received: 10/10 replicas passed. Gate passed.
• quality_gate_idle, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_idle, bounds check intake_connections: 10/10 replicas passed. Gate passed.
• quality_gate_security_idle, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_security_idle, bounds check cpu_usage: 10/10 replicas passed. Gate passed.
• quality_gate_metrics_logs, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_metrics_logs, bounds check missed_bytes: 10/10 replicas passed. Gate passed.
• quality_gate_metrics_logs, bounds check cpu_usage: 10/10 replicas passed. Gate passed.
• quality_gate_metrics_logs, bounds check intake_connections: 10/10 replicas passed. Gate passed.
• quality_gate_metrics_logs, bounds check total_bytes_received: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check intake_connections: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check total_bytes_received: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check missed_bytes: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_security_mean_fs_load, bounds check cpu_usage: 10/10 replicas passed. Gate passed.
• quality_gate_security_mean_fs_load, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_idle_all_features, bounds check intake_connections: 10/10 replicas passed. Gate passed.
• quality_gate_idle_all_features, bounds check total_bytes_received: 10/10 replicas passed. Gate passed.
• quality_gate_idle_all_features, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_security_no_fs_load, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_security_no_fs_load, bounds check cpu_usage: 10/10 replicas passed. Gate passed.

[atanzu]

This might probably be a separate structure instead - less error-prone (because of field names). According to the ABI doc performance shall stay the same.

[vickenty]

Unfortunately, returning a struct seems to be a bit slower than a list of primitives by about 2% across architectures. I've added benchmarks to the PR description.

[RiantZ]

This is not part of current PR, but it attracts my attention: the retry/payload-split sentinel now travels back out through dist.WriteTo(pb). It works for today's implementers, but any future Distribution that wraps the error (%w) silently breaks payload splitting → dropped sketch data. Shall we use errors.Is(err, errRetry)?

[vickenty]

Good point. I don't think there is much point two wrap anything in WriteTo, and errors.Is look more expensive than the simple compare. I think I'll clarify this in WriteTo docs so it's not hidden.

[RiantZ]

Shall we move numPoints to DDSketchPoints as Len() method?

[RiantZ]

We are passing to compareFn parameters as Actual, Expected.
While at line 97 of current file:

assertSketchSeriesEqualWithComparator(t, exp, act, func(exp, act *quantile.Sketch) bool {
		return quantile.SketchesApproxEqual(exp, act, e)
	})

Expected, Actual! It strange that tests are passed.

[vickenty]

Well spotted. This is a frequent bug that usually only affects failure formatting, since most comparison operations are symmetric, and do not care about order of arguments.

[RiantZ]

Methods for Otel histograms will be part of another PR?
I'm asking because this PR is a preparation step for OTEL histogram support. Perhaps having these methods in place (in current PR) would help us assess whether we are moving in the right direction?

[vickenty]

I think the Otel methods should be added when we actually implement them, to avoid dead code.

The API would look pretty similar:

type DistributionWriter interface {
    //...
    WriteOtelExplicitBucketsHistogram(meta, numPoints, points OtelExplicitBucketsHistogramPoints) error
}

type OtelExplicitBucketHistogramPoints interface {
    GetOtelExplicitBucketsHistogramPoint(int) (timestamp, bounds, counts, ...)
}

[StephenWakely]

Looks good!