performance: Reduce allocations (logger factory reuse, check enabled)

[RicoSuter]

Proposed changes

Some performance optimizations, reducing memory allocations by ~23% while maintaining full API compatibility and thread-safety.

Related Issues

• Fixes Possible memory allocation regression in v1.5.378.10-preview #3349

Types of changes

What types of changes does your code introduce?

• Bugfix (non-breaking change which fixes an issue)
• Enhancement (non-breaking change which adds functionality)
• Test enhancement (non-breaking change to increase test coverage)
• Breaking change (fix or feature that would cause existing functionality to not work as expected, requires version increase of Nuget packages)
• Documentation Update (if none of the other choices apply)

Checklist

Put an x in the boxes that apply. You can also fill these out after creating the PR. If you're unsure about any of them, don't hesitate to ask. We're here to help! This is simply a reminder of what we are going to look for before merging your code.

• I have read the CONTRIBUTING doc.
• I have signed the CLA.
• I ran tests locally with my changes, all passed.
• I fixed all failing tests in the CI pipelines.
• I fixed all introduced issues with CodeQL and LGTM.
• I have added tests that prove my fix is effective or that my feature works and increased code coverage.
• I have added necessary documentation (if appropriate).
• Any dependent changes have been merged and published in downstream modules.

Further comments

1. LoggerFactory Singleton Pattern

File: Stack/Opc.Ua.Core/Types/Utils/LoggerUtils.cs:95-99

Introduced singleton pattern for TraceLoggerTelemetry to eliminate repeated allocations:

private static readonly TraceLoggerTelemetry s_sharedTelemetry = new();

static Utils()
{
    TelemetryExtensions.InternalOnly__TelemetryHook = () => s_sharedTelemetry;
}

Impact: Primary source of 23% allocation reduction
Safety: Thread-safe, no API changes

2. Additional Optimizations

The upstream library also includes:

• ConcurrentDictionary for lock-free reads in Subscription.cs
• Enhanced EventSource/ETW telemetry
• Queue handling improvements
• MonitoredItem optimizations

All changes maintain backward compatibility.

Benchmark results

Master branch (baseline):

===========================================================================================================================================
Client Benchmark - 1 minute - [2025-12-03 11:22:29.342]

Total received changes:          1194276
Total published changes:         1000000
Process memory:                  555.02 MB (289.25 MB in .NET heap)
Avg allocations over last 60s:   97.03 MB/s

Metric                               Avg        P50        P90        P95        P99      P99.9        Max        Min     StdDev      Count
-------------------------------------------------------------------------------------------------------------------------------------------
Received (changes/s)            19917.67   19902.19   21523.12   22004.03   23617.96   23617.96   23617.96   15474.36    1398.62          -
Processing latency (ms)             4.78       2.37      14.25      20.25      29.60      39.71      73.16       0.00       6.52    1194276
End-to-end latency (ms)            82.79      71.80     146.49     165.29     270.05     387.09     427.97       0.90      51.32    1194276
===========================================================================================================================================
Server Benchmark - 1 minute - [2025-12-03 11:22:29.764]

Total received changes:          1006334
Total published changes:         1192400
Process memory:                  718.87 MB (448.94 MB in .NET heap)
Avg allocations over last 60s:   505.28 MB/s

Metric                               Avg        P50        P90        P95        P99      P99.9        Max        Min     StdDev      Count
-------------------------------------------------------------------------------------------------------------------------------------------
Received (changes/s)            16737.91   16894.72   17389.25   17503.45   17549.66   17549.66   17549.66   13848.97     690.37          -
Processing latency (ms)             0.00       0.00       0.00       0.01       0.01       0.07     244.58       0.00       0.25    1006334
End-to-end latency (ms)           332.08     332.24     546.72     591.58     713.86     802.54     941.21       0.77     167.16    1006334

This PR (optimizations):

===========================================================================================================================================
Client Benchmark - 1 minute - [2025-12-03 11:08:41.825]

Total received changes:          1194087
Total published changes:         1007994
Process memory:                  545.32 MB (290.39 MB in .NET heap)
Avg allocations over last 60s:   74.04 MB/s

Metric                               Avg        P50        P90        P95        P99      P99.9        Max        Min     StdDev      Count
-------------------------------------------------------------------------------------------------------------------------------------------
Received (changes/s)            19903.96   19906.61   21330.26   21587.95   21664.65   21664.65   21664.65   16468.73    1118.32          -
Processing latency (ms)             5.78       2.71      18.04      25.35      35.88      44.56     142.66       0.01       8.35    1194087
End-to-end latency (ms)            76.47      67.54     137.84     154.21     191.12     283.48     314.49       1.19      44.27    1194087
===========================================================================================================================================
Server Benchmark - 1 minute - [2025-12-03 11:08:43.428]

Total received changes:          1000896
Total published changes:         1193530
Process memory:                  737.14 MB (423.26 MB in .NET heap)
Avg allocations over last 60s:   474.08 MB/s

Metric                               Avg        P50        P90        P95        P99      P99.9        Max        Min     StdDev      Count
-------------------------------------------------------------------------------------------------------------------------------------------
Received (changes/s)            16662.03   16716.28   17198.44   17389.35   17589.10   17589.10   17589.10   14610.73     573.32          -
Processing latency (ms)             0.00       0.00       0.00       0.01       0.01       0.08      23.24       0.00       0.05    1000896
End-to-end latency (ms)           302.90     300.90     504.83     544.12     665.81     736.51     760.37       0.21     153.72    1000896

Client => from 97 mb/s to 74 mb/s, 5ms lower latency on avg
Server => from 505 mb/s to 474 mb/s, 30ms lower latency on avg

[CLAassistant]

All committers have signed the CLA.

[RicoSuter]

Test Project	Status	Results
Opc.Ua.Types.Tests	✅ PASSED	288 tests passed
Opc.Ua.Security.Certificates.Tests	✅ PASSED	144 tests passed
Opc.Ua.Encoders.Fuzz.Tests	✅ PASSED	150 tests passed
Opc.Ua.Configuration.Tests	✅ PASSED	All tests passed
Opc.Ua.PubSub.Tests	✅ PASSED	All tests passed
Opc.Ua.Core.Tests	✅ PASSED	All tests passed
Opc.Ua.Server.Tests	✅ PASSED	All tests passed
Opc.Ua.Gds.Tests	✅ PASSED	All tests passed
Opc.Ua.Client.Tests	✅ PASSED	Failed: 0, Passed: 1001, Skipped: 163
Opc.Ua.Client.ComplexTypes.Tests	✅ PASSED	Failed: 0, Passed: 3574, Skipped: 0

[codecov]

Codecov Report

❌ Patch coverage is 11.11111% with 64 lines in your changes missing coverage. Please review.
✅ Project coverage is 59.95%. Comparing base (d25caff) to head (f89a12f).
⚠️ Report is 26 commits behind head on master.

Files with missing lines	Patch %	Lines
...raries/Opc.Ua.Client/Subscription/MonitoredItem.cs	15.78%	14 Missing and 2 partials ⚠️
...ies/Opc.Ua.Server/Server/OpcUaServerEventSource.cs	0.00%	13 Missing and 2 partials ⚠️
.../Opc.Ua.Server/Subscription/SubscriptionManager.cs	0.00%	10 Missing and 3 partials ⚠️
Libraries/Opc.Ua.Client/OpcUaClientEventSource.cs	0.00%	5 Missing and 3 partials ⚠️
...braries/Opc.Ua.Client/Subscription/Subscription.cs	27.27%	6 Missing and 2 partials ⚠️
...nitoredItem/QueueHandler/DataChangeQueueHandler.cs	0.00%	1 Missing and 3 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3387      +/-   ##
==========================================
+ Coverage   51.86%   59.95%   +8.08%     
==========================================
  Files         370      374       +4     
  Lines       78618    78304     -314     
  Branches    13650    13620      -30     
==========================================
+ Hits        40779    46947    +6168     
+ Misses      33705    27026    -6679     
- Partials     4134     4331     +197     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[marcschier]

Thanks @RicoSuter - this is great, one more thing off my plate :-)

[RicoSuter]

@marcschier i think the changes here are really low-risk, no? Reusing the factory instead of creating needs to be checked but seems also safe for me... will dig deeper into more complex improvements when i have time.

[RicoSuter]

@marcschier my personal goal is still to achieve 1.2 mio changes in both directions and on server side <100ms average :-) not even sure whether this is possible with OPC UA :-D

[marcschier]

@RicoSuter, one of the changes we need for that to happen is to reduce GC pressure for monitored item notification and DataValue, Variant, and node id allocations on the subscribe/publish paths client side, we have some benchmarks and the improvements are worth it although it will break apis. Goal is before April next year.

[marcschier]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[RicoSuter]

@marcschier do you expect that I add unit tests for the changed LoCs?

[marcschier]

@RicoSuter - No, most of it was for legacy pieces, and the one you fixed is actively tested already... Sorry took longer to merge the queue is quite full right now...

[RicoSuter]

@marcschier I've run benchmarks again and updated the PR:

Client receive processing looks good now (even a bit faster), server side still degraded (4-8%) but this is less a problem because most clients do not write that much.

Comparison grid here:
RicoSuter/Namotion.Interceptor#104

Thanks for merging my PR.