enhancement: batch timestamped (passthrough) metrics for a short period of time before forwarding #426

tobz · 2025-01-15T18:04:39Z

Summary

This PR adds the ability for the aggregate transform to batch "passthrough" (pre-aggregated) metrics for short periods of time, in larger-than-normal event buffers, with the express goal of improving the efficiency of handling pre-aggregated metrics.

We've updated the logic of the transform to follow an equivalent behavior in the Datadog Agent's "no aggregation pipeline", which looks something like:

batch pre-aggregated metrics into their own buffer (defaults to 2048 points)
if the batch is filled to capacity, flush it to the "forwarder" (Datadog Metrics destination, in our case)
every two seconds, check if we have received any pre-aggregated metrics within the last second: if not, flush the batch to the "forwarder"

This aims to improve how many pre-aggregated metrics are packed into an individual series/sketch request, improving efficiency and reducing the number of requests that have to be sent. There's still a difference in number of series/sketch requests sent between Core Agent and ADP even with this batching behavior in place, which I'm still currently investigating in staging.

Change Type

Bug fix
New feature
Non-functional (chore, refactoring, docs)
Performance

How did you test this PR?

Tested in staging.

(more detail to be added here as I test further)

References

N/A

pr-commenter · 2025-01-15T18:34:59Z

Regression Detector (Saluki)

Regression Detector Results

Run ID: a76a760d-897d-444d-a3c0-d7247cb38421

Baseline: 2b44efa
Comparison: 44da86b
Diff

Optimization Goals: ✅ No significant changes detected

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI	trials
➖	dsd_uds_500mb_3k_contexts	ingress throughput	+0.81	[+0.69, +0.93]	1
➖	dsd_uds_1mb_50k_contexts_memlimit	ingress throughput	+0.74	[+0.21, +1.28]	1
➖	dsd_uds_100mb_3k_contexts_distributions_only	memory utilization	+0.43	[+0.31, +0.55]	1
➖	dsd_uds_10mb_3k_contexts	ingress throughput	+0.01	[-0.02, +0.04]	1
➖	dsd_uds_100mb_3k_contexts	ingress throughput	+0.01	[-0.04, +0.06]	1
➖	dsd_uds_50mb_10k_contexts_no_inlining	ingress throughput	+0.01	[-0.06, +0.07]	1
➖	dsd_uds_40mb_12k_contexts_40_senders	ingress throughput	+0.00	[-0.02, +0.03]	1
➖	dsd_uds_50mb_10k_contexts_no_inlining_no_allocs	ingress throughput	+0.00	[-0.05, +0.05]	1
➖	dsd_uds_1mb_3k_contexts	ingress throughput	+0.00	[-0.02, +0.02]	1
➖	dsd_uds_1mb_50k_contexts	ingress throughput	-0.00	[-0.00, +0.00]	1
➖	dsd_uds_1mb_3k_contexts_dualship	ingress throughput	-0.00	[-0.00, +0.00]	1
➖	dsd_uds_512kb_3k_contexts	ingress throughput	-0.01	[-0.02, +0.01]	1
➖	dsd_uds_100mb_250k_contexts	ingress throughput	-0.01	[-0.04, +0.03]	1
➖	quality_gates_idle_rss	memory utilization	-0.59	[-0.62, -0.56]	1

Bounds Checks: ✅ Passed

perf	experiment	bounds_check_name	replicates_passed	links
✅	quality_gates_idle_rss	memory_usage	10/10

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:

Its estimated |Δ mean %| ≥ 5.00%, indicating the change is big enough to merit a closer look.
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.
Its configuration does not mark it "erratic".

pr-commenter · 2025-01-15T18:36:09Z

Regression Detector (DogStatsD)

Regression Detector Results

Run ID: 6778acac-bd47-4ea7-aa8b-6e9ace904ed9

Baseline: 7.63.0-rc.2
Comparison: 7.63.0-rc.2

Optimization Goals: ✅ No significant changes detected

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI	trials
➖	dsd_uds_40mb_12k_contexts_40_senders	ingress throughput	+0.00	[-0.00, +0.00]	1
➖	dsd_uds_1mb_50k_contexts	ingress throughput	+0.00	[-0.00, +0.00]	1
➖	dsd_uds_512kb_3k_contexts	ingress throughput	+0.00	[-0.01, +0.01]	1
➖	dsd_uds_1mb_50k_contexts_memlimit	ingress throughput	+0.00	[-0.00, +0.00]	1
➖	dsd_uds_1mb_3k_contexts	ingress throughput	-0.00	[-0.00, +0.00]	1
➖	dsd_uds_1mb_3k_contexts_dualship	ingress throughput	-0.00	[-0.00, +0.00]	1
➖	dsd_uds_100mb_250k_contexts	ingress throughput	-0.00	[-0.00, +0.00]	1
➖	dsd_uds_10mb_3k_contexts	ingress throughput	-0.00	[-0.00, +0.00]	1
➖	dsd_uds_100mb_3k_contexts	ingress throughput	-0.01	[-0.05, +0.04]	1
➖	dsd_uds_100mb_3k_contexts_distributions_only	memory utilization	-0.32	[-0.46, -0.17]	1
➖	dsd_uds_500mb_3k_contexts	ingress throughput	-2.08	[-2.22, -1.94]	1
➖	quality_gates_idle_rss	memory utilization	-2.10	[-2.22, -1.99]	1

Bounds Checks: ❌ Failed

perf	experiment	bounds_check_name	replicates_passed	links
❌	quality_gates_idle_rss	memory_usage	0/10

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:

Its estimated |Δ mean %| ≥ 5.00%, indicating the change is big enough to merit a closer look.
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.
Its configuration does not mark it "erratic".

pr-commenter · 2025-01-15T18:36:53Z

Regression Detector Links

Experiment Result Links

experiment	link(s)
dsd_uds_100mb_250k_contexts	[Profiling (ADP)] [Profiling (DSD)] [SMP Dashboard]
dsd_uds_100mb_3k_contexts	[Profiling (ADP)] [Profiling (DSD)] [SMP Dashboard]
dsd_uds_100mb_3k_contexts_distributions_only	[Profiling (ADP)] [Profiling (DSD)] [SMP Dashboard]
dsd_uds_10mb_3k_contexts	[Profiling (ADP)] [Profiling (DSD)] [SMP Dashboard]
dsd_uds_1mb_3k_contexts	[Profiling (ADP)] [Profiling (DSD)] [SMP Dashboard]
dsd_uds_1mb_3k_contexts_dualship	[Profiling (ADP)] [Profiling (DSD)] [SMP Dashboard]
dsd_uds_1mb_50k_contexts	[Profiling (ADP)] [Profiling (DSD)] [SMP Dashboard]
dsd_uds_1mb_50k_contexts_memlimit	[Profiling (ADP)] [Profiling (DSD)] [SMP Dashboard]
dsd_uds_40mb_12k_contexts_40_senders	[Profiling (ADP)] [Profiling (DSD)] [SMP Dashboard]
dsd_uds_500mb_3k_contexts	[Profiling (ADP)] [Profiling (DSD)] [SMP Dashboard]
dsd_uds_512kb_3k_contexts	[Profiling (ADP)] [Profiling (DSD)] [SMP Dashboard]
quality_gates_idle_rss	[Profiling (ADP)] [Profiling (DSD)] [SMP Dashboard]
dsd_uds_50mb_10k_contexts_no_inlining (ADP only)	[Profiling (ADP)] [SMP Dashboard]
dsd_uds_50mb_10k_contexts_no_inlining_no_allocs (ADP only)	[Profiling (ADP)] [SMP Dashboard]

…od of time before forwarding

rayz · 2025-02-05T16:31:16Z

lib/saluki-components/src/transforms/aggregate/mod.rs

+            self.forward_events(forwarder).await;
+
+            if self.active_buffer.try_push(event).is_some() {
+                error!("Event buffer is full even after forwarding events. Dropping event.");


Do we need to add a return here to prevent line 490 from hitting?

Yeah, adding a return here makes sense. 👍🏻

Fixed in 75b6f96.

I also now increment the "events dropped" metric to reflect the reflect we're legitimately dropping a metric on the floor in that branch.

… is full

github-actions bot added area/components Sources, transforms, and destinations. transform/aggregate Aggregate transform. labels Jan 15, 2025

tobz force-pushed the tobz/add-batching-timestamped-metrics-agg branch 4 times, most recently from af4d048 to 3dc98ab Compare January 17, 2025 18:37

tobz added 5 commits February 3, 2025 18:18

enhancement: batch timestamped (passthrough) metrics for a short peri…

5a50562

…od of time before forwarding

use larger, dedicated event buffers for no-agg data flow

cd17945

really, truly try to mimic the Agent's no agg behavior

de2e245

track how long each passthrough batch is alive

731c1d1

push passthrough logic into dedicated type + better telemetry

f76068d

tobz force-pushed the tobz/add-batching-timestamped-metrics-agg branch from 49022d1 to f76068d Compare February 3, 2025 23:21

slight cleanup

44da86b

tobz marked this pull request as ready for review February 4, 2025 20:56

tobz requested a review from a team as a code owner February 4, 2025 20:56

rayz reviewed Feb 5, 2025

View reviewed changes

rayz previously approved these changes Feb 5, 2025

View reviewed changes

return early + increment events dropped when replacement event buffer…

75b6f96

… is full

tobz dismissed rayz’s stale review via 75b6f96 February 5, 2025 18:32

tobz requested a review from rayz February 5, 2025 18:33

rayz approved these changes Feb 5, 2025

View reviewed changes

tobz merged commit bf76762 into main Feb 5, 2025
20 checks passed

tobz deleted the tobz/add-batching-timestamped-metrics-agg branch February 5, 2025 19:03

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

enhancement: batch timestamped (passthrough) metrics for a short period of time before forwarding #426

enhancement: batch timestamped (passthrough) metrics for a short period of time before forwarding #426

tobz commented Jan 15, 2025 •

edited

Loading

pr-commenter bot commented Jan 15, 2025 •

edited

Loading

Fine details of change detection per experiment

Bounds Checks: ✅ Passed

Explanation

pr-commenter bot commented Jan 15, 2025 •

edited

Loading

Fine details of change detection per experiment

Explanation

pr-commenter bot commented Jan 15, 2025 •

edited

Loading

rayz Feb 5, 2025

tobz Feb 5, 2025

tobz Feb 5, 2025

enhancement: batch timestamped (passthrough) metrics for a short period of time before forwarding #426

enhancement: batch timestamped (passthrough) metrics for a short period of time before forwarding #426

Conversation

tobz commented Jan 15, 2025 • edited Loading

Summary

Change Type

How did you test this PR?

References

pr-commenter bot commented Jan 15, 2025 • edited Loading

Regression Detector (Saluki)

Regression Detector Results

Optimization Goals: ✅ No significant changes detected

Fine details of change detection per experiment

Bounds Checks: ✅ Passed

Explanation

pr-commenter bot commented Jan 15, 2025 • edited Loading

Regression Detector (DogStatsD)

Regression Detector Results

Optimization Goals: ✅ No significant changes detected

Fine details of change detection per experiment

Bounds Checks: ❌ Failed

Explanation

pr-commenter bot commented Jan 15, 2025 • edited Loading

Regression Detector Links

Experiment Result Links

rayz Feb 5, 2025

Choose a reason for hiding this comment

tobz Feb 5, 2025

Choose a reason for hiding this comment

tobz Feb 5, 2025

Choose a reason for hiding this comment

tobz commented Jan 15, 2025 •

edited

Loading

pr-commenter bot commented Jan 15, 2025 •

edited

Loading

pr-commenter bot commented Jan 15, 2025 •

edited

Loading

pr-commenter bot commented Jan 15, 2025 •

edited

Loading