Logging Benchmarks

[Aaronontheweb]

Changes

The Akka.NET logging system fundamentals haven't really changed much since 2014 or so, and I suspect we are doing some extremely ugly things in that code. Adding some benchmarks using a custom ILoggingAdapter implementation so I can measure the overhead.

Checklist

For significant changes, please ensure that the following have been completed (delete if not relevant):

• This change follows the Akka.NET API Compatibility Guidelines.
• This change follows the Akka.NET Wire Compatibility Guidelines.
• I have reviewed my own pull request.

Latest dev Benchmarks

BenchmarkDotNet=v0.13.2, OS=Windows 10 (10.0.19043.2006/21H1/May2021Update)
Intel Core i7-8850H CPU 2.60GHz (Coffee Lake), 1 CPU, 12 logical and 6 physical cores
.NET SDK=6.0.100
  [Host]     : .NET 6.0.9 (6.0.922.41905), X64 RyuJIT AVX2
  Job-JJNOTA : .NET 6.0.9 (6.0.922.41905), X64 RyuJIT AVX2

InvocationCount=1  UnrollFactor=1  

Method	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Gen1	Gen2	Allocated	Alloc Ratio
LogInfoNoParams	159.0 ns	3.18 ns	6.42 ns	1.00	0.00	0.0110	0.0040	0.0010	64 B	1.00
LogInfo1Params	663.5 ns	11.04 ns	9.79 ns	4.15	0.20	0.0280	0.0110	0.0030	160 B	2.50
LogInfo2Params	849.3 ns	14.42 ns	12.78 ns	5.31	0.29	0.0340	0.0140	0.0040	192 B	3.00
LogInfoWithException	162.7 ns	3.11 ns	5.99 ns	1.03	0.05	0.0110	0.0040	0.0010	64 B	1.00
LogInfoWithException1Params	669.4 ns	13.36 ns	12.50 ns	4.20	0.21	0.0280	0.0110	0.0030	160 B	2.50
LogInfoWithException2Params	850.4 ns	11.39 ns	10.10 ns	5.32	0.25	0.0340	0.0140	0.0040	192 B	3.00

Bear in mind - these results are scaled down per operation. This means writing a single log event in Akka.NET is roughly 10x more expensive than processing a message by default.