Comparing MRP performance with embedded advanced scheduling - impact on planning cycle time

[john_tech]

I wanted to start a discussion about performance differences between classic MRP batch processing and embedded PP/DS advanced scheduling in S/4HANA 1809. We’re evaluating whether to migrate from our current MRP Live approach to embedded advanced scheduling for better constraint-based planning.

Our current MRP batch runs process 85,000 materials across 8 plants nightly, taking approximately 4.5 hours. The planning results are generally good, but we lack finite capacity scheduling and can’t model complex resource constraints. We’re considering embedded PP/DS, but I’m concerned about performance and scalability given the significantly more complex calculations involved.

Has anyone made this transition? What was your experience with planning run duration compared to classic MRP? Did the improved planning accuracy justify any performance trade-offs? I’m particularly interested in understanding whether the embedded approach can scale to our material volume while maintaining acceptable planning cycle times for daily replanning scenarios.

[john_tech]

Let me synthesize the key considerations for your evaluation based on practical implementation experience:

Classic MRP Batch vs. Embedded Advanced Scheduling Performance: The performance comparison isn’t straightforward because you’re comparing different planning paradigms. Classic MRP uses simpler infinite capacity logic with material availability checks, while embedded PP/DS adds finite capacity scheduling, detailed resource modeling, and constraint-based optimization. Your current 4.5-hour MRP runtime will likely increase to 6-8 hours initially with full PP/DS implementation. However, this isn’t an apples-to-apples comparison - you’re gaining significantly more sophisticated planning capabilities.

The performance differential comes from:

• Resource capacity checks for every operation (not just material availability)
• Setup time optimization and sequence-dependent scheduling
• Alternative routing evaluation for capacity leveling
• Real-time constraint propagation across the planning network

To minimize performance impact, implement these strategies:

1. Use net change planning for daily runs instead of regenerative - reduces planning scope by 70-80%
2. Implement planning time fences to stabilize near-term schedules and limit replanning
3. Configure appropriate planning horizons per material group (don’t plan 52 weeks for fast-moving items)
4. Use the hybrid finite/infinite approach mentioned earlier for non-bottleneck resources

Performance and Scalability Concerns: Scalability to 85,000 materials is definitely achievable, but requires proper system sizing and optimization. Key scalability factors:

1. Hardware Requirements: Embedded PP/DS is more memory-intensive than MRP. For your volume, ensure minimum 768GB HANA RAM with sufficient CPU cores (48+ recommended). The planning engine is highly parallelizable, so multi-core performance matters more than clock speed.

2. Master Data Quality: Poor master data kills PP/DS performance. Ensure:

   • Accurate production versions and routing validity dates
   • Properly maintained work center capacities and calendars
   • Clean BOM structures without circular dependencies
   • Realistic lot sizes and minimum order quantities

3. Planning Segmentation: Don’t run all 85,000 materials in one monolithic planning run. Segment by:

   • Product families with independent supply chains
   • Planning areas (plant/storage location combinations)
   • ABC classification (critical A items with finite scheduling, C items with simpler logic) This allows parallel planning execution and faster cycle times.

4. Technical Optimization: Enable these PP/DS-specific optimizations:

   • Use planning heuristic SAP_PP_ALG_002 (optimized for S/4HANA 1809)
   • Configure resource planning table partitioning for large work centers
   • Implement incremental deployment to reduce memory footprint
   • Enable parallel processing with appropriate server groups

Impact on Planning Accuracy: This is where embedded PP/DS truly delivers value that justifies performance trade-offs:

1. Finite Capacity Scheduling: You’ll identify capacity overloads before they become expediting emergencies. Our implementations typically see 15-25% reduction in schedule breaks and expediting costs.

2. Resource Constraint Modeling: Ability to model setup times, tool availability, skill requirements, and alternative work centers enables realistic schedules. Planning accuracy (measured as planned vs. actual completion) improves by 10-15 percentage points on average.

3. Optimization vs. Feasibility: Classic MRP generates material plans that might not be executable due to capacity constraints. PP/DS generates executable schedules from the start, reducing planner intervention time by 30-40%.

Recommendation for Your Scenario: Given your 85,000 material volume and 8-plant complexity, I’d recommend a phased approach:

Phase 1 (Months 1-3): Implement embedded PP/DS for 1-2 pilot plants with your most constrained resources. Accept 50-75% longer planning times initially. Focus on proving the planning accuracy improvements.

Phase 2 (Months 4-6): Optimize performance based on pilot learnings. Target 20-30% reduction from initial runtimes through tuning. Expand to additional plants.

Phase 3 (Months 7-9): Full rollout with hybrid finite/infinite approach. Target final planning cycle time of 5-6 hours for daily net change runs, 8-10 hours for weekly regenerative runs.

The performance trade-off is real but manageable, and the planning accuracy improvements typically justify it for complex manufacturing environments with genuine capacity constraints. If your environment has simple routings, abundant capacity, and primarily material-constrained planning, classic MRP might be sufficient and faster.

[j_osat]

We migrated a similar-sized operation from MRP to embedded PP/DS last year. Initial planning runs were significantly slower - about 8-9 hours for full regenerative planning. However, after optimization (proper master data setup, planning horizons, and using net change planning instead of regenerative), we got it down to 5-6 hours. The key benefit wasn’t speed but planning quality. Our schedule attainment improved from 78% to 91% because of finite capacity scheduling. Sometimes slower but more accurate planning is worth it.

[kevin_admin]

Performance in embedded PP/DS is heavily dependent on your resource and capacity model complexity. If you’re modeling every single work center with detailed capacity constraints, calendars, and setup matrices, it will be slower than MRP. But you can use a hybrid approach - run finite scheduling only for bottleneck resources and use infinite planning for non-constraints. This gives you 80% of the benefit with much better performance. We process 120,000 materials in about 6 hours using this approach.

[carol_data]

That’s a helpful perspective on the hybrid approach. How do you determine which resources are bottlenecks programmatically? Is there a way to dynamically adjust which work centers get finite scheduling based on capacity utilization, or do you manually designate them in the master data? Also, can you switch between finite and infinite mode without full replanning?

[mary_expert]

Another consideration is the technical architecture. Embedded PP/DS leverages HANA’s in-memory processing, so performance scales with your hardware. We saw a 40% improvement in planning runtime after upgrading from 512GB to 1TB RAM and adding CPU cores. If you’re resource-constrained on the HANA side, PP/DS will struggle. Also, make sure you’re using the optimized planning heuristics introduced in 1809 - the legacy algorithms are much slower.

[patricia_expert]

You can use capacity evaluation reports to identify bottlenecks and set the planning mode per resource in the master data. The planning heuristic respects these settings. For dynamic adjustment, you’d need to implement custom logic that updates resource master data based on utilization thresholds, but that’s fairly complex. Most organizations identify their 5-10 critical bottleneck resources and keep those settings stable. Switching modes does require replanning for affected materials.

[kungfudev90]

Don’t overlook the impact on planning accuracy though. We stayed with classic MRP because our production environment has so much variability that the detailed capacity constraints in PP/DS were constantly violated anyway. The planning runs took twice as long and the schedules weren’t much more achievable. If your shop floor has high schedule adherence and low variability, PP/DS makes sense. If not, the performance hit might not be justified.