WorkflowForge Internal Benchmarks

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

Version: 2.1.1
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: 50 iterations per benchmark, 5 warmup iterations
Last Updated: March 6, 2026

Table of Contents

Executive Summary

Median numbers (50 iterations):

Metric Result
Operation Execution 8.8-82μs median (excluding delays)
Operation Creation 1.2-1.9μs median
Workflow Throughput 35-272μs for custom operations (1-50 ops, all runtimes)
Memory Baseline 3,408 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
82μs
Max CPU-bound Op
3.3KB
Minimal Footprint
15.9x
Concurrency Speedup
1.9μs
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 12.1μs 10.9μs 8.8μs 1,912 B
ConditionalOperationFalse 34.3μs 33.2μs 29.2μs 1,072 B
ConditionalOperationTrue 34.7μs 33.2μs 31.6μs 1,016 B
CustomOperationExecution 33.7μs 33.8μs 27.9μs 456 B
DelegateOperationExecution 33.2μs 42.6μs 29.0μs 616 B
ActionOperationExecution 42.6μs 42.1μs 29.9μs 648 B
ForEachSmallCollection 44.7μs 49.1μs 32.7μs 2,336 B
ForEachLargeCollection 63.5μs 70.2μs 50.2μs 7,128 B
WithRestoration 50.7μs 45.5μs 29.7μs 568 B
DataManipulation 61.1μs 55.1μs 64.6μs 8,536 B
ChainedOperations 79.5μs 78.0μs 54.9μs 4,720 B
ExceptionHandling 81.7μs 59.3μs 66.3μs 2,944 B
DelayOperationExecution 15,192μs 15,132μs 15,353μs 1,432 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 1.9μs 1.6μs 1.6μs 56 B
ActionCreation 1.9μs 1.7μs 1.3μs 56 B
CustomCreation 1.7μs 1.4μs 1.2μs 32 B
Operation Execution Times (Median, Lower is Better)
CPU-bound ops stay under 82μs median in this slice
10.9μs
42.6μs
33.8μs
42.1μs
49.1μs
70.2μs
78μs
59.3μs
.NET 10.0
12.1μs
33.2μs
33.7μs
42.6μs
44.7μs
63.5μs
79.5μs
81.7μs
.NET 8.0
8.8μs
29μs
27.9μs
29.9μs
32.7μs
50.2μs
54.9μs
66.3μs
.NET FX 4.8
WorkflowForge (Logging → Delegate → Custom → Action → ForEach S → ForEach L → Chained → Exception)

Observations

  • Custom ops remain the cheapest allocation (456 B here).
  • Logging ops are fastest (8.8–12.1μs).
  • Construction costs ~1.2–1.9μs median.
  • Exception handling is visible; .NET 10.0 trims it (59.3μs vs 81.7μs 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 50.6μs 47.6μs 34.6μs 3,096 B CPU-bound
HighPerformanceConfiguration 53.6μs 52.2μs 33.0μs 3,672 B CPU-bound
ForEachLoopWorkflow 60.8μs 59.2μs 39.0μs 5,176 B CPU-bound
SequentialDelegateOperations 15,047μs 15,086μs 15,313μs 3,880 B Delay-bound
DataPassingWorkflow 15,066μs 15,068μs 15,354μs 4,024 B Delay-bound
ConditionalOperationsWorkflow 15,136μs 15,096μs 15,271μs 4,352 B Delay-bound
LoggingOperationsWorkflow 15,059μs 15,115μs 15,302μs 6,184 B Delay-bound
MemoryIntensiveWorkflow 15,036μs 15,170μs 15,331μs 4,856 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 50.6μs 47.6μs 34.6μs 3,096 B
5 74.4μs 72.0μs 67.3μs 7,120 B
10 88.6μs 88.3μs 99.1μs 12,192 B
25 148.7μs 150.2μs 154.0μs 29,688 B
50 215.3μs 211.1μs 264.3μs 59,328 B
Custom Operation Throughput Scaling (1-50 Operations)
50 custom ops finish under 272μs median; memory rises linearly with op count
47.6μs
72μs
88.3μs
150.2μs
211.1μs
.NET 10.0
50.6μs
74.4μs
88.6μs
148.7μs
215.3μs
.NET 8.0
34.6μs
67.3μs
99.1μs
154μs
264.3μs
.NET FX 4.8
SequentialCustomOperations (1→5→10→25→50 ops)

Numbers:

  • CPU-bound workflows finish in about 35–272μs 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 32.2μs 34.6μs 32.2μs 3,408 B No Gen0/1/2
SmallObjectAllocation 101.2μs 93.3μs 88.6μs 19,928 B —
StringConcatenationAllocation 97.5μs 90.1μs 87.9μs 17,904 B —
StringBuilderOptimization 93.0μs 96.5μs 93.8μs 17,264 B —
CollectionAllocation 97.7μs 101.3μs 81.2μs 18,912 B —
ObjectPoolingSimulation 124.8μs 120.5μs 90.6μs 23,000 B —
ArrayReuseOptimization 121.6μs 120.2μs 114.0μs 23,248 B —
MemoryPressureScenario 241.9μs 224.1μs 190.9μs 319,024 B —
LargeObjectAllocation 664.1μs 608.3μs 622.9μs 1,019,608 B Gen0+Gen1+Gen2
DisposableResourceManagement 159,520μs 159,606μs 160,583μs 20,784 B Delay-bound

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 3,408 B 3,408 B
50 3,408 B 3,408 B
100 3,408 B 3,408 B
500 3,408 B 3,408 B

The minimal allocation workflow holds a flat 3,408 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 ~3.3KB; large-object path stresses Gen2. .NET FX 4.8 skips alloc metrics.
3.3KB
17.3KB
19.9KB
23.2KB
319KB
1.0MB
.NET 10.0
3.3KB
17.3KB
19.9KB
23.2KB
319KB
1.0MB
.NET 8.0
Typical (no GC pressure)
Elevated
High (Gen0/1/2)

Observations

  • Still 3,408 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 ~640 B vs raw concatenation at ten passes (17,264 B vs 17,904 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 (456 B per execution).
  2. Logging operations when the step is tiny; they lead the timing table at 8.8–12.1μs.
  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 3,408 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.