Mini Projeto 09 - Benchmark

Benchmark project comparing EF Core (tracking and no-tracking) vs Dapper using BenchmarkDotNet and PostgreSQL.

Description

This repository provides a reproducible benchmarking harness to compare data access performance between EF Core and Dapper under equivalent query scenarios. It focuses on read-heavy workloads and highlights the trade-offs between productivity and raw speed.

Features

• Benchmark scenarios: ById, PagedList, FilterByStatus
• EF Core with tracking and AsNoTracking
• Dapper queries for the same scenarios
• Dataset seeding with configurable size
• BenchmarkDotNet configuration (warmup, iterations, exporters)
• Docker Compose for local PostgreSQL

Tech Stack

• .NET 10
• BenchmarkDotNet
• EF Core 10 + Npgsql
• Dapper
• PostgreSQL 16

Error Contract

This project is a console benchmark tool, not an HTTP API. There is no error contract or response schema.

Endpoints

None. This project does not expose HTTP endpoints.

How to Run

1) Start PostgreSQL (Docker)

docker compose up -d

The compose file exposes PostgreSQL on port 5433.

2) Set connection string

$env:BENCHMARK_DB = "Host=localhost;Port=5433;Database=benchmarks;Username=postgres;Password=postgres"

3) Seed dataset

dotnet run --project .\benchmarks\DataAccess.Benchmarks -- --seed --count 50000 --reset

4) Run benchmarks (Release)

dotnet run -c Release --project .\benchmarks\DataAccess.Benchmarks -- --filter *

PowerShell shortcuts

.\scripts\seed.ps1 -Count 50000 -Reset
.\scripts\bench.ps1 -Filter * -Min 10000 -Configuration Release

Examples

Run a smaller dataset:

.\scripts\seed.ps1 -Count 10000 -Reset

Run only the ById benchmarks:

.\scripts\bench.ps1 -Filter *ById* -Configuration Release

Benchmark Results (Cycle 2)

Environment (2026-02-02)

• Windows 11 25H2 (10.0.26200.7623)
• AMD Ryzen 7 6800H (8c/16t)
• .NET SDK 10.0.102 / Runtime 10.0.2
• PostgreSQL 16 (Docker, port 5433)
• Dataset: 50,000 rows

Parameters used in summary

• Paged: PageSize=50, PageNumber=5
• Status: Status=2, PageSize=50

Scenario	ORM	Mean (ms)	StdDev	Alloc (KB)	Difference (%)
ById	EF Core (Tracking)	0.743	0.086	40.66	+48.3
ById	EF Core (NoTracking)	0.622	0.018	40.71	+24.2
ById	Dapper	0.501	0.026	1.68	0
Paged	EF Core (Tracking)	0.672	0.036	86.05	+33.6
Paged	EF Core (NoTracking)	0.640	0.044	58.96	+27.2
Paged	Dapper	0.503	0.020	12.47	0
Status	EF Core (Tracking)	0.716	0.049	87.21	+31.0
Status	EF Core (NoTracking)	0.893	0.174	60.08	+63.3
Status	Dapper	0.547	0.040	12.48	0

Cycle 1 vs Cycle 2 (Percent Change)

Change in Mean from Cycle 1 to Cycle 2 (positive = slower).

Scenario	ORM	Mean Cycle 1 (ms)	Mean Cycle 2 (ms)	Change (%)
ById	EF Core (Tracking)	0.575	0.743	+29.2
ById	EF Core (NoTracking)	0.561	0.622	+10.9
ById	Dapper	0.484	0.501	+3.5
Paged	EF Core (Tracking)	0.749	0.672	-10.3
Paged	EF Core (NoTracking)	0.740	0.640	-13.5
Paged	Dapper	0.511	0.503	-1.6
Status	EF Core (Tracking)	0.731	0.716	-2.1
Status	EF Core (NoTracking)	0.731	0.893	+22.2
Status	Dapper	0.518	0.547	+5.6

Conclusion

There is no absolute rule. Choose the tool based on latency goals, workload volume, and team productivity.