Skip to content

performance.md

Latest commit

 

History

History
88 lines (66 loc) · 4.04 KB

performance.md

File metadata and controls

88 lines (66 loc) · 4.04 KB
Raw

Performance Benchmarks

Overview

Benchmarks are run with BenchmarkDotNet comparing Fake4Dataverse (Lenient preset) against FakeXrmEasy 1.x on .NET Framework 4.6.2 and FakeXrmEasy 3.x on .NET 10.

Important notes:

  • FakeXrmEasy 1.x targets .NET Framework only, so it can only be compared on net462.
  • FakeXrmEasy 3.x (FakeXrmEasy.Core.v9) is benchmarked with the RPL_1_5 license and minimal middleware setup (.AddCrud().UseCrud()).
  • Fake4Dataverse uses the Lenient preset (all auto-behaviors off) for an apples-to-apples comparison.

Environment

Detail Value
OS Windows 11
CPU Intel Core Ultra 7 258V
BenchmarkDotNet v0.14.0
Job ShortRun (3 warmup iterations, 3 target iterations)

CRUD Operations

Single-entity operations on an account record.

Benchmark F4D (net462) FXE v1 (net462) Speedup F4D (net10) FXE v3 (net10) Speedup
Create 2.54 µs 10.43 µs 4.1× 2.62 µs 14.2 µs 5.4×
Create + Update + Delete 1.28 µs 16.95 µs 13.3× 744 ns 29.8 µs 40×
Retrieve (by ID) 213 ns 9.68 µs 45× 136 ns 13.3 µs 98×
Update 353 ns 5.48 µs 15.5× 168 ns 7.75 µs 46×

QueryExpression / RetrieveMultiple

Paged query returning page 1 with 50 records per page.

Scenario Rows F4D (net462) FXE v1 (net462) Speedup F4D (net10) FXE v3 (net10) Speedup
All columns 100 59.9 µs 1.69 ms 28× 40.3 µs 2.10 ms 52×
Filter (statecode=0) 100 79.2 µs 2.73 ms 34× 54.6 µs 3.17 ms 58×
Order by name 100 141 µs 4.54 ms 32× 66.5 µs 2.17 ms 33×
Filter + order 100 134 µs 5.66 ms 42× 62.5 µs 3.27 ms 52×
All columns 1,000 781 µs 15.32 ms 19.6× 353 µs 7.83 ms 22×
Filter (statecode=0) 1,000 1.22 ms 13.83 ms 11.3× 518 µs 9.26 ms 18×
Order by name 1,000 2.38 ms 19.04 ms 8.0× 1.11 ms 10.1 ms 9.1×
Filter + order 1,000 1.56 ms 16.19 ms 10.4× 813 µs 9.60 ms 11.8×
All columns 10,000 25.3 ms 221.6 ms 8.8× 12.3 ms 117 ms 9.5×
Filter (statecode=0) 10,000 26.3 ms 163.0 ms 6.2× 13.4 ms 91.6 ms 6.8×
Order by name 10,000 42.3 ms 235.1 ms 5.6× 16.0 ms 116 ms 7.2×
Filter + order 10,000 33.6 ms 200.6 ms 6.0× 16.3 ms 97.1 ms 6.0×

Key Takeaways

  • Fake4Dataverse is 4–98× faster than FakeXrmEasy across all operations.
  • Retrieve by ID shows the largest difference (45–98× faster) because Fake4Dataverse uses a direct dictionary lookup with no query-pipeline overhead.
  • .NET 10 is consistently faster than .NET Framework 4.6.2 for both libraries, but the gap is larger for Fake4Dataverse due to heavy use of Span<T> and modern collection APIs.
  • At 10,000 rows the speedup narrows (5–10×) because both libraries are dominated by the same O(n) scan; adding an AttributeIndex in Fake4Dataverse can restore sub-millisecond filtered queries even at large row counts.

Running Benchmarks Yourself

The benchmark project lives in benchmarks/Fake4Dataverse.Benchmarks.

# .NET Framework 4.6.2 — vs FakeXrmEasy 1.x
dotnet run --configuration Release --project benchmarks/Fake4Dataverse.Benchmarks -f net462

# .NET 10 — vs FakeXrmEasy 3.x
dotnet run --configuration Release --project benchmarks/Fake4Dataverse.Benchmarks -f net10.0

Results are written to BenchmarkDotNet.Artifacts/results/.

Tip: Use --filter *Crud* or --filter *Query* to run a subset of benchmarks.

Optimization Tips

For guidance on improving query performance in your tests — including attribute indexes, column-set reduction, and paging strategies — see the Performance & Indexing guide.