[.NET Core] Event Source traces revision - Part 3

[cheenamalhotra]

Managed SNI layer full tracing coverage from TdsParser.

[Wraith2]

Before releasing this can we make sure we do a memory profile and see if we're causing any new allocations from the tracing? These add to some very hot paths and could have high performance costs.

[David-Engel]

Partial review...

[DavoudEshtehari]

Using a constant string to address a class method instead of using the strong types makes it vulnerable to future maintenance and possible misinformation. I suggest avoiding this approach at least hereafter.

[cheenamalhotra]

@DavoudEshtehari

Please checkout latest commit to avoid human errors in method/class names, let me know if you want to propose a change before I go forward to rest of files.

[DavoudEshtehari]

Here are the perf report for Managed SNI on Windows without enabling the traces by Switch.Microsoft.Data.SqlClient.UseManagedNetworkingOnWindows=true:

• Microsoft.Data.SqlClient V2.1.1:

BenchmarkDotNet=v0.12.1, OS=Windows 10.0.19042
Intel Core i7-8665U CPU 1.90GHz (Coffee Lake), 1 CPU, 8 logical and 4 physical cores
.NET Core SDK=5.0.103
  [Host]     : .NET Core 3.1.12 (CoreCLR 4.700.21.6504, CoreFX 4.700.21.6905), X64 RyuJIT
  Job-EEOJRM : .NET Core 3.1.12 (CoreCLR 4.700.21.6504, CoreFX 4.700.21.6905), X64 RyuJIT

IterationCount=10  LaunchCount=1  WarmupCount=10  

Method	Mean	Error	StdDev	Gen 0	Gen 1	Gen 2	Allocated	Completed Work Items	Lock Contentions
Connection_Open	8.236 μs	0.1842 μs	0.1096 μs	0.2747	0.0916	-	1.13 KB	0.0000	-
Connection_OpenAsync	9.885 μs	2.5592 μs	1.6927 μs	0.3204	0.1068	-	1.32 KB	0.0000	-
ExecuteScalar	25,271.486 μs	2,088.2995 μs	1,242.7135 μs	406.2500	406.2500	406.2500	20482.32 KB	0.0625	-
ExecuteScalarAsync	3,478,298.100 μs	411,327.5428 μs	272,067.7018 μs	3000.0000	2000.0000	1000.0000	31193.54 KB	1320.0000	-
ExecuteNonQuery	17,598.683 μs	1,469.9404 μs	972.2745 μs	-	-	-	1.81 KB	0.0625	-
ExecuteNonQueryAsync	18,371.153 μs	931.5047 μs	616.1327 μs	-	-	-	3.44 KB	3.0625	-
ExecuteReader	29,459.933 μs	2,726.4595 μs	1,622.4723 μs	406.2500	406.2500	406.2500	20482.6 KB	0.0625	-
ExecuteReaderAsync	3,225,492.210 μs	485,161.6103 μs	320,904.3659 μs	3000.0000	2000.0000	1000.0000	31193.24 KB	1318.0000	-
ExecuteXmlReader	102,757.710 μs	3,769.9511 μs	2,493.5892 μs	7000.0000	4000.0000	1800.0000	82081.99 KB	0.4000	-
ExecuteXmlReaderAsync	114,564.100 μs	6,156.8096 μs	3,220.1305 μs	5200.0000	2800.0000	800.0000	82156.11 KB	3.4000	-

• Current change from here:

BenchmarkDotNet=v0.12.1, OS=Windows 10.0.19042
Intel Core i7-8665U CPU 1.90GHz (Coffee Lake), 1 CPU, 8 logical and 4 physical cores
.NET Core SDK=5.0.103
  [Host]     : .NET Core 3.1.12 (CoreCLR 4.700.21.6504, CoreFX 4.700.21.6905), X64 RyuJIT
  Job-FSZQZJ : .NET Core 3.1.12 (CoreCLR 4.700.21.6504, CoreFX 4.700.21.6905), X64 RyuJIT

IterationCount=10  LaunchCount=1  WarmupCount=10  

Method	Mean	Error	StdDev	Gen 0	Gen 1	Gen 2	Allocated	Completed Work Items	Lock Contentions
Connection_Open	7.535 μs	0.5108 μs	0.2672 μs	0.2747	0.0916	-	1.13 KB	0.0000	-
Connection_OpenAsync	7.358 μs	0.8022 μs	0.4196 μs	0.3204	0.1068	-	1.32 KB	0.0000	-
ExecuteScalar	25,644.887 μs	3,266.8449 μs	2,160.8157 μs	406.2500	406.2500	406.2500	20482.88 KB	0.0625	-
ExecuteScalarAsync	3,110,033.890 μs	386,661.8388 μs	255,752.8658 μs	3000.0000	2000.0000	1000.0000	31194.66 KB	1320.0000	-
ExecuteNonQuery	14,510.836 μs	239.7518 μs	158.5810 μs	-	-	-	2.42 KB	0.0625	-
ExecuteNonQueryAsync	15,998.277 μs	329.1947 μs	195.8985 μs	-	-	-	4.28 KB	3.0625	-
ExecuteReader	30,085.312 μs	3,241.0842 μs	1,695.1497 μs	406.2500	406.2500	406.2500	20483.16 KB	0.0625	-
ExecuteReaderAsync	2,936,545.167 μs	304,153.8034 μs	180,997.0497 μs	3000.0000	2000.0000	1000.0000	31194.09 KB	1318.0000	-
ExecuteXmlReader	128,351.173 μs	19,679.6004 μs	13,016.8372 μs	7000.0000	4000.0000	1750.0000	82082.99 KB	0.5000	-
ExecuteXmlReaderAsync	135,127.420 μs	42,969.2720 μs	28,421.5130 μs	5500.0000	3250.0000	1000.0000	82158.5 KB	3.5000	-

[Wraith2]

LGTM