WorkflowForge Internal Benchmarks

WorkflowForge-only BenchmarkDotNet results: per-operation cost, throughput sweeps, memory, concurrency. No competitor mix-ins.

Version: 2.2.0
Test System: Windows 11 (25H2), Intel 11th Gen i7-1185G7, .NET SDK 10.0.103
Runtimes: .NET 10.0.3, .NET 8.0.24, .NET Framework 4.8.1
Benchmark Framework: BenchmarkDotNet v0.15.8
Methodology: 10 iterations per benchmark job
Last Updated: September 15, 2026

Table of Contents

Executive Summary

Median numbers (10 iterations per job):

Metric Result
Operation Execution 73.6ns–9.5μs median (excluding delays)
Operation Creation 48.1ns median
Workflow Throughput ~4–211μs for custom operations (1–50 ops, .NET 8/10)
Memory Baseline 4,126 B minimal allocation (constant)
Concurrency Scaling Roughly linear (8.0x for 8 workers, 15.9x for 16 workers)
GC Pressure Gen0 only for typical workloads
33.9μs
Max CPU-bound Op
4.03KB
Minimal Footprint
15.9x
Concurrency Speedup
48.1ns
Op Creation

Operation Performance

Per-operation timings and allocations on .NET 8.0, 10.0, and .NET Framework 4.8.

Operation Execution (Median Times)

Operation Type .NET 8.0 .NET 10.0 .NET FX 4.8 Allocated (.NET 8)
LoggingOperationExecution 76.2ns 73.6ns 216.2ns 136 B
ConditionalOperationFalse 1.9μs 1.9μs 6.1μs 1,008 B
ConditionalOperationTrue 1.9μs 1.9μs 6.1μs 952 B
CustomOperationExecution 1.3μs 1.3μs 3.5μs 392 B
DelegateOperationExecution 1.4μs 1.5μs 3.9μs 552 B
ActionOperationExecution 1.5μs 1.5μs 4.0μs 584 B
ForEachSmallCollection 2.6μs 2.5μs 7.5μs 2,279 B
ForEachLargeCollection 7.1μs 4.7μs 32.3μs 7,073 B
WithRestoration 2.1μs 2.1μs 7.8μs 568 B
DataManipulation 23.9μs 15.5μs 57.1μs 8,536 B
ChainedOperations 10.0μs 9.5μs 32.3μs 5,378 B
ExceptionHandling 33.9μs 16.6μs 43.6μs 2,560 B
DelayOperationExecution 16.1ms 16.1ms 16.1ms 752 B

DelayOperationExecution contains a 1ms delay; .NET FX 4.8 does not report allocation metrics.

Operation Creation (Median Times)

Operation Type .NET 8.0 .NET 10.0 .NET FX 4.8 Allocated (.NET 8)
DelegateCreation 51.2ns 51.0ns 67.9ns 56 B
ActionCreation 51.2ns 51.0ns 67.9ns 56 B
CustomCreation 50.6ns 48.1ns 59.7ns 32 B
Operation Execution Times (Median, Lower is Better)
CPU-bound ops stay under 43.6μs median in this slice
73.6ns
1.5μs
1.3μs
1.5μs
2.5μs
4.7μs
9.5μs
16.6μs
.NET 10.0
76.2ns
1.4μs
1.3μs
1.5μs
2.6μs
7.1μs
10.0μs
33.9μs
.NET 8.0
216.2ns
3.9μs
3.5μs
4.0μs
7.5μs
32.3μs
32.3μs
43.6μs
.NET FX 4.8
WorkflowForge (Logging → Delegate → Custom → Action → ForEach S → ForEach L → Chained → Exception)

Observations

  • Custom ops remain the cheapest allocation (392 B here).
  • Logging ops are fastest (73.6ns–216.2ns).
  • Construction costs ~48–68ns median.
  • Exception handling is visible; .NET 10.0 trims it (16.6μs vs 33.9μs on .NET 8.0 for the same benchmark).

Workflow Throughput

End-to-end workflow shapes with increasing operation counts. Delay-heavy cases still include the intentional ~1ms sleeps.

Workflow Patterns (OperationCount=1, Median by Runtime)

Pattern .NET 8.0 .NET 10.0 .NET FX 4.8 Memory (.NET 8) Notes
SequentialCustomOperations 4.1μs 4.0μs 8.7μs 3,819 B CPU-bound
HighPerformanceConfiguration 4.6μs 4.3μs 11.2μs 4,392 B CPU-bound
ForEachLoopWorkflow 6.3μs 5.9μs 15.1μs 5,888 B CPU-bound
SequentialDelegateOperations 15.9ms 15.9ms 15.9ms 4,587 B Delay-bound
DataPassingWorkflow 15.9ms 15.9ms 15.9ms 4,741 B Delay-bound
ConditionalOperationsWorkflow 15.9ms 15.9ms 15.9ms 5,079 B Delay-bound
LoggingOperationsWorkflow 15.9ms 15.9ms 15.9ms 5,068 B Delay-bound
MemoryIntensiveWorkflow 15.9ms 15.9ms 15.8ms 5,591 B Delay-bound

Memory column shows .NET 8.0 allocation; .NET FX 4.8 allocation metrics are NA.

Throughput Scaling (SequentialCustomOperations, Median by Runtime)

Operations .NET 8.0 .NET 10.0 .NET FX 4.8 Memory (.NET 8)
1 4.1μs 4.0μs 8.7μs 3,819 B
5 34.9μs 32.2μs 65.1μs 7,761 B
10 50.2μs 47.6μs 99.0μs 12,677 B
25 57.1μs 53.1μs 175.7μs 29,685 B
50 109.8μs 98.2μs 320.8μs 58,521 B
Custom Operation Throughput Scaling (1-50 Operations)
50 custom ops finish under 321μs median; memory rises linearly with op count
4.0μs
32.2μs
47.6μs
53.1μs
98.2μs
.NET 10.0
4.1μs
34.9μs
50.2μs
57.1μs
109.8μs
.NET 8.0
8.7μs
65.1μs
99.0μs
175.7μs
320.8μs
.NET FX 4.8
SequentialCustomOperations (1→5→10→25→50 ops)

Numbers:

  • CPU-bound workflows finish in about 4–321μs for 1–50 operations across the tested runtimes.
  • Allocations grow in step with operation count in the sequential custom-op sweep.
  • Delay-heavy rows are dominated by the sleeps, not the orchestration loop.

Memory Allocation

Allocation and GC behavior (median over 10 allocations unless noted).

Allocation Patterns (10 Allocations, Median by Runtime)

Pattern .NET 8.0 .NET 10.0 .NET FX 4.8 Memory (.NET 8) GC
MinimalAllocationWorkflow 3.5μs 3.4μs 7.6μs 4,126 B —
SmallObjectAllocation 49.3μs 44.0μs 105.6μs 20,398 B —
StringConcatenationAllocation 48.2μs 42.1μs 102.1μs 18,370 B —
StringBuilderOptimization 45.3μs 40.8μs 101.2μs 17,489 B —
CollectionAllocation 49.7μs 46.2μs 108.7μs 19,384 B —
ObjectPoolingSimulation 57.0μs 52.7μs 115.9μs 23,040 B —
ArrayReuseOptimization 35.6μs 30.9μs 139.4μs 24,381 B —
MemoryPressureScenario 195.6μs 175.0μs 243.9μs 319,518 B —
LargeObjectAllocation 591.3μs 521.1μs 694.0μs 1,020,293 B —
DisposableResourceManagement 160.8ms 160.7ms 161.2ms 21,278 B —

Memory column shows .NET 8.0 allocation; .NET Framework 4.8 allocation metrics are NA (Allocated column shows “—” in benchmarks).

Memory Scaling (MinimalAllocationWorkflow, .NET 8.0)

Allocations .NET 8.0 Memory .NET 10.0 Memory
10 4,126 B 4,147 B
50 4,126 B 4,147 B
100 4,126 B 4,147 B
500 4,126 B 4,147 B

The minimal allocation workflow holds a flat 4,126 B across 10–500 allocations in this benchmark, which matches the table above. .NET Framework 4.8 does not report allocation metrics here.

Memory Allocation Patterns (10 Allocations)
Minimal workflow holds ~4.03KB; large-object path stresses Gen2. .NET FX 4.8 skips alloc metrics.
4.05KB
17.11KB
19.94KB
23.83KB
312KB
996.4KB
.NET 10.0
4.03KB
17.08KB
19.92KB
23.81KB
312KB
996.38KB
.NET 8.0
Typical (no GC pressure)
Elevated
High (Gen0/1/2)

Observations

  • Still 4,126 B from 10 through 500 iterations in the minimal row.
  • Typical patterns skipped GC pressure in the minimal scenario.
  • LargeObjectAllocation hits Gen0+1+2 as expected.
  • StringBuilderOptimization saves ~881 B vs raw concatenation at ten passes (17,489 B vs 18,370 B).

Concurrency Scaling

Concurrent workflow fan-out versus sequential baselines on .NET 8.0, 10.0, and .NET Framework 4.8.

Scaling with 8 Workflows (5 ops per workflow)

Pattern .NET 8.0 .NET 10.0 .NET FX 4.8 Memory (.NET 8) Speedup
SequentialWorkflows 626.1 ms 629.2 ms 641.6 ms 86.25 KB —
ConcurrentWorkflows 78.9 ms 78.8 ms 79.2 ms 89.25 KB 8.0x
ParallelWorkflows 78.8 ms 78.8 ms 79.2 ms 89.2 KB 8.0x

Scaling with 16 Workflows (5 ops per workflow)

Pattern .NET 8.0 .NET 10.0 .NET FX 4.8 Memory (.NET 8) Speedup
SequentialWorkflows 1,239 ms 1,255 ms 1,289 ms 172.21 KB —
ConcurrentWorkflows 79.0 ms 79.1 ms 79.6 ms 178.09 KB 15.9x

Scaling with 8 Workflows, 25 ops each

Pattern .NET 8.0 .NET 10.0 .NET FX 4.8 Memory (.NET 8) Speedup
SequentialWorkflows 3,130 ms 3,147 ms 3,221 ms 411.13 KB —
ConcurrentWorkflows 399.2 ms 400.1 ms 400.9 ms 414.13 KB 7.8x
Concurrency Scaling (Sequential vs Concurrent Time)
8 workflows ≈8.0x, 16 workflows ≈15.9x vs sequential in this harness
629ms
79ms
.NET 10.0 (8 wf, 8.0x)
1.26s
79ms
.NET 10.0 (16 wf, 15.9x)
626ms
79ms
.NET 8.0 (8 wf, 8.0x)
1.24s
79ms
.NET 8.0 (16 wf, 15.7x)
642ms
79ms
.NET FX 4.8 (8 wf, 8.1x)
1.29s
80ms
.NET FX 4.8 (16 wf, 16.2x)
Sequential Execution
Concurrent Execution

What matters:

  • Measured speedups land at 8.0x for eight workflows and 15.9x for sixteen versus sequential baselines in this harness.
  • Reported memory rises in proportion to workflow count.
  • With 25 operations per workflow, eight concurrent workers still reach about 7.8x versus sequential.

Optimization Recommendations

  1. Custom operations for the smallest allocations in this matrix (392 B per execution).
  2. Logging operations when the step is tiny; they lead the timing table at 73.6ns–216.2ns.
  3. Avoid LOH churn in hot loops; big allocations invite Gen2 pauses.
  4. Parallelize deliberately; scaling here is roughly linear (8.0x for 8 workers, 15.9x for 16).
  5. Reuse buffers and properties; MinimalAllocationWorkflow flatlines at 4,126 B from 10 through 500 iterations.
  6. Use .NET 10.0 where you can; several ops and exception paths improve vs .NET 8.0 in the same harness.