WorkflowForge Competitive Benchmark Analysis

Analysis Date: January 2026
Frameworks Tested:

• WorkflowForge 2.0.0
• Workflow Core
• Elsa Workflows

Test System: Windows 11 (25H2), Intel 11th Gen i7-1185G7, .NET 8.0.23
BenchmarkDotNet: v0.15.8

Table of Contents

• Executive Summary
• Scenario Breakdown
  • Scenario 1: Simple Sequential Workflow
  • Scenario 2: Data Passing Workflow
  • Scenario 3: Conditional Branching
  • Scenario 4: Loop/ForEach Processing
  • Scenario 5: Concurrent Execution
  • Scenario 6: Error Handling
  • Scenario 7: Creation Overhead
  • Scenario 8: Complete Lifecycle
  • Scenario 9: State Machine
  • Scenario 10: Long Running
  • Scenario 11: Parallel Execution
  • Scenario 12: Event-Driven
• Performance Advantage Summary
• Architectural Differences
• Benchmark Methodology
• Statistical Significance
• Conclusion

Executive Summary

WorkflowForge demonstrates 11-540x faster execution and 9-573x less memory allocation compared to Workflow Core and Elsa Workflows across 12 real-world scenarios (50 iterations per benchmark).

Key Insights:

• WorkflowForge operates at microsecond scale (13-497μs), competitors at millisecond scale (0.8-94ms)
• Memory allocations remain in kilobytes (3.5-121KB) vs. megabytes (0.04-19MB) for competitors
• State Machine scenarios show highest advantage: up to 540x faster vs Elsa
• Concurrent Execution shows 109-264x faster performance
• Sequential Workflows show 26-71x faster with minimal memory
• Consistent performance across all 12 scenario types

Visual Performance Comparison

Execution Time (Lower is Better)

Memory Allocation (Lower is Better)

Scaling Charts - Performance Advantage Grows with Workload

Key Finding: WorkflowForge’s advantage increases with workload size.

Scenario Breakdown

Scenario 1: Simple Sequential Workflow

Description: Execute operations sequentially (1, 5, 10, 25, 50 operations)

Performance Results (Median Times, 50 iterations)

Memory Allocation Results

Key Insight: WorkflowForge performance advantage increases linearly with operation count.

Scenario 2: Data Passing Workflow

Description: Pass data between operations (5, 10, 25 operations)

Performance Results (Median Times, 50 iterations)

Memory Allocation Results

Key Insight: Data passing overhead is minimal in WorkflowForge (<1μs per operation).

Scenario 3: Conditional Branching

Description: Conditional logic with if/else branches (10, 25, 50 operations)

Performance Results (Median Times)

Key Insight: Conditional overhead negligible (<1μs per branch decision).

Scenario 4: Loop/ForEach Processing

Description: Iterate over collections (10, 50, 100 items)

Performance Results (Median Times)

Memory Allocation Results

Key Insight: ForEach performance advantage increases with collection size.

Scenario 5: Concurrent Execution

Description: Execute multiple workflows concurrently (1, 4, 8 workflows)

Performance Results (Median Times)

Memory Allocation Results

Key Insight: WorkflowForge maintains consistent per-workflow overhead (~16KB) regardless of concurrency.

Scenario 6: Error Handling

Description: Exception handling and recovery

Performance Results (Median Times)

Advantage:

• 11.1x faster than Workflow Core
• 64.4x faster than Elsa

Memory: WorkflowForge 5KB vs. Workflow Core 46KB vs. Elsa 1,072KB

Key Insight: Error handling overhead is minimal (~111μs) in WorkflowForge.

Scenario 7: Creation Overhead

Description: Workflow instantiation cost

Performance Results (Median Times)

Advantage:

• 63x faster than Workflow Core
• 162x faster than Elsa

Memory: WorkflowForge 3.73KB vs. Workflow Core 129KB vs. Elsa 578KB

Key Insight: WorkflowForge workflow creation is negligible (~13μs).

Scenario 8: Complete Lifecycle

Description: Full create-execute-dispose cycle

Performance Results (Median Times)

Advantage:

• 236x faster than Elsa
• Workflow Core excluded due to architectural incompatibility (see note below)

Memory: WorkflowForge 3.57KB vs. Elsa 1,510KB (423x less)

Note: Workflow Core was excluded from this benchmark due to an architectural design difference. Workflow Core’s WorkflowHost.Start() method spins up background worker threads that are intended to run continuously, making rapid create-start-stop-dispose cycles (50 iterations) incompatible with its design. This is a fundamental architectural difference, not a performance issue. For more details, see the benchmark documentation.

Key Insight: Complete lifecycle overhead is trivial (~42μs) in WorkflowForge.

Scenario 9: State Machine

Description: State machine with multiple transitions (5, 10, 25 transitions)

Performance Results (Median Times)

Memory Allocation Results

Key Insight: State machine execution shows the highest performance advantage (up to 540x faster).

Scenario 10: Long Running

Description: Long-running operations with delays (delay-bound scenario)

Performance Results (Median Times)

Memory Allocation Results (5 ops/1ms delay)

Advantage: Similar timing (delay-bound), but 51x less memory than WC, 423x less than Elsa.

Key Insight: Long-running workflows are delay-bound; advantage is in memory efficiency.

Scenario 11: Parallel Execution

Description: Parallel operation execution within a workflow (4, 8, 16 operations)

Performance Results (Median Times)

Memory Allocation Results (16 ops/4 concurrency)

Key Insight: Parallel execution maintains 38-380x speed advantage with 15-573x less memory.

Scenario 12: Event-Driven

Description: Event-driven workflow execution with delays

Performance Results (Median Times)

Memory Allocation Results

Advantage: Similar timing to WorkflowCore, but 11x less memory than WC, 296x less than Elsa.

Key Insight: Event-driven scenarios are I/O-bound; advantage is in memory efficiency.

Performance Advantage Summary

By Scenario Type (12 Scenarios)

Overall Speed Range: 11-540x faster execution (compute-bound scenarios)
Overall Memory Range: 9-573x less memory allocation

Key Findings

1. State Machine scenarios show the highest speed advantage: 303-540x faster
2. Concurrent Execution maintains excellent scaling: 109-264x faster
3. Long Running and Event-Driven are I/O-bound, but memory savings are massive: 51-423x less
4. Memory efficiency is consistently excellent across all 12 scenarios

Architectural Differences

WorkflowForge Design

1. Lightweight Execution Model
   • No background threads
   • Synchronous sequential execution (default)
   • Explicit parallelism via ForEachWorkflowOperation
   • Minimal object allocations
2. Dictionary-Based Data Flow
   • ConcurrentDictionary<string, object?> for properties
   • Zero serialization overhead
   • Thread-safe property access
3. Dependency-Free Core
   • No reflection-heavy frameworks
   • No serialization frameworks
   • Pure .NET Standard 2.0
4. Middleware Pipeline
   • Russian Doll pattern
   • Minimal delegate allocations
   • No reflection per operation

Workflow Core Design

1. Persistent Workflow Engine
   • Background worker threads
   • Designed for long-running workflows
   • Persistent state management
   • Work queue architecture
2. Strong Typing
   • Reflection-based step resolution
   • JSON serialization for state
3. Intended Use Case
   • Long-running business processes (hours/days)
   • Workflows that survive process restarts
   • Background processing

Conclusion: Workflow Core is optimized for durability, not speed. WorkflowForge is optimized for speed, not durability (though persistence extensions available).

Elsa Workflows Design

1. Workflow Designer Focus
   • Visual workflow designer
   • HTTP workflow triggers
   • Extensive activity library
2. Serialization-Heavy
   • JSON serialization for all data
   • Heavy use of reflection
   • Large object graphs
3. Intended Use Case
   • Visual workflow design
   • Human task workflows
   • Integration workflows

Conclusion: Elsa is optimized for designer experience, not performance. WorkflowForge is optimized for performance, not visual design (though programmatic API is very expressive).

Benchmark Methodology

Test Configuration

• BenchmarkDotNet: v0.15.8
• Runtime: .NET 8.0.23
• Iterations: 50 per benchmark
• Warmup: 5 iterations
• Invocation: 1 per iteration
• Unroll Factor: 1

Hardware

• OS: Windows 11 (10.0.26200.6899)
• CPU: Intel 11th Gen i7-1185G7
• Memory: Sufficient for all benchmarks

Scenario Implementations

All scenarios implement identical logic across all frameworks:

• Same operation count
• Same data structures
• Same conditional logic
• Same collection sizes
• Same concurrency levels

Fairness Verification:

• Workflow Core implementations use TaskCompletionSource for precise completion detection
• Elsa implementations use proper workflow completion await
• WorkflowForge implementations use standard ForgeAsync()

Reproduction

Full benchmark source code available in repository:

• src/benchmarks/WorkflowForge.Benchmarks.Comparative/
• All scenarios in Scenarios/ folder
• Run via dotnet run -c Release

Statistical Significance

All results meet statistical significance criteria:

• Standard deviation < 20% of mean (most scenarios)
• P95 values show consistency
• Median values used for comparison (more stable than mean)
• 50 iterations provide statistical confidence

Outliers: Some scenarios show high standard deviation due to GC pauses or system activity. Median values are used to minimize impact.

Conclusion

WorkflowForge delivers 11-540x faster execution and 9-573x less memory allocation compared to Workflow Core and Elsa Workflows across 12 scenarios. This performance advantage stems from:

1. Architectural Simplicity: No background threads, no persistent state, no serialization
2. Minimal Allocations: Dictionary-based data flow, no reflection per operation
3. Optimized Execution: Russian Doll middleware, efficient delegate handling
4. Focused Design: Optimized for speed, not durability or visual design

WorkflowForge is the fastest .NET workflow engine for high-performance, programmatic workflow orchestration.

References

• WorkflowForge: https://github.com/animatlabs/workflow-forge
• Workflow Core: https://github.com/danielgerlag/workflow-core
• Elsa Workflows: https://github.com/elsa-workflows/elsa-core

Related Documentation

• Performance Overview - Internal WorkflowForge benchmarks
• Architecture Overview - Design principles
• Getting Started - Quick start guide