Skip to content

benchmark.md

Latest commit

 

History

History
120 lines (80 loc) · 3.58 KB

benchmark.md

File metadata and controls

120 lines (80 loc) · 3.58 KB

Benchmarks

The benchmark project (FirebirdSql.Data.FirebirdClient.Benchmarks) measures performance of the Firebird .NET provider using BenchmarkDotNet.

Prerequisites

  • A running Firebird server accessible at localhost (default configuration)
  • .NET 10 SDK

Running Benchmarks

Use the convenience script from the repository root:

.\run-benchmark.ps1

By default this runs CommandBenchmark. To select a different benchmark class:

.\run-benchmark.ps1 -Benchmark ConnectionBenchmark
.\run-benchmark.ps1 -Benchmark LargeFetchBenchmark

Advanced Options

Enable JIT disassembly output:

.\run-benchmark.ps1 -Disasm

Enable ETW profiling (Windows only):

.\run-benchmark.ps1 -Profile

Options can be combined:

.\run-benchmark.ps1 -Benchmark LargeFetchBenchmark -Disasm

Running Directly with BenchmarkDotNet

For full control over BenchmarkDotNet options, pass arguments directly after --:

dotnet run --project src\FirebirdSql.Data.FirebirdClient.Benchmarks\FirebirdSql.Data.FirebirdClient.Benchmarks.csproj --configuration Release -- --list flat
dotnet run --project src\FirebirdSql.Data.FirebirdClient.Benchmarks\FirebirdSql.Data.FirebirdClient.Benchmarks.csproj --configuration Release -- --filter "*Fetch*"

Connection String

By default the benchmark connects to:

database=localhost:benchmark.fdb;user=sysdba;password=masterkey

Override this with the FIREBIRD_BENCHMARK_CS environment variable:

$env:FIREBIRD_BENCHMARK_CS = "database=myhost:benchmark.fdb;user=sysdba;password=masterkey"
.\run-benchmark.ps1

Configuration

All benchmarks share a common configuration (BenchmarkConfig):

  • Baseline job: ReleaseNuGet build configuration — references the latest published FirebirdSql.Data.FirebirdClient NuGet package.
  • Candidate job: Release build configuration — references the local project source.
  • Runtime: .NET 10
  • Diagnostics: Memory allocations (MemoryDiagnoser)
  • Export: GitHub-flavored Markdown table (written to BenchmarkDotNet.Artifacts/)
  • Ordering: Fastest to slowest

The NuGet baseline lets you compare the locally built provider against the published release to detect regressions or measure improvements.

Available Benchmarks

CommandBenchmark

Measures command execution over two data types (BIGINT, VARCHAR(10) CHARACTER SET UTF8):

Benchmark Description
Execute / ExecuteAsync Inserts 100 rows using ExecuteNonQuery / ExecuteNonQueryAsync
Fetch / FetchAsync Reads 100 rows using ExecuteReader / ExecuteReaderAsync

ConnectionBenchmark

Measures connection pool throughput:

Benchmark Description
OpenClose / OpenCloseAsync Opens and closes a pooled connection

LargeFetchBenchmark

Measures bulk read throughput for 100,000 rows across five data types:

Data Type Notes
BIGINT Fixed-size integer
CHAR(255) CHARACTER SET UTF8 Fixed-length string
CHAR(255) CHARACTER SET OCTETS Fixed-length binary
BLOB SUB_TYPE TEXT CHARACTER SET UTF8 Text blob
BLOB SUB_TYPE BINARY Binary blob

Results

BenchmarkDotNet writes results to BenchmarkDotNet.Artifacts/ in the repository root. This directory is listed in .gitignore. Each run produces:

  • A summary table in the console
  • A GitHub-flavored Markdown file (.md) suitable for pasting into issues or pull requests
  • An HTML report
  • CSV data