Skip to content

Finding Bottlenecks

Identify what's limiting your system's performance.

Goal

By the end of this tutorial, you will:

  • Identify system bottlenecks systematically
  • Use metrics to pinpoint constraints
  • Understand different bottleneck types
  • Plan interventions effectively

Time: 30-45 minutes

Prerequisites

Step 1: Run and Collect Data

Run a simulation with detailed metrics:

waremax run scenario.yaml \
  -o results/ \
  --metrics detailed

Step 2: Quick Bottleneck Scan

Use the analyze command:

waremax analyze results/ --bottlenecks

Output:

=== Bottleneck Analysis ===

Primary Bottleneck: Station S1
  - Utilization: 95%
  - Avg queue: 6.2 robots
  - Throughput limited by service capacity

Secondary Bottleneck: Edge E15
  - Occupancy: 88%
  - Traffic congestion causing delays

Impact:
  - 23% of task time spent waiting for S1
  - 15% of task time in traffic delays

Step 3: Identify Bottleneck Type

Station Bottleneck

Symptoms: 

Station utilization > 90%
Long station queues
Low overall throughput despite available robots

Verification: 

waremax analyze results/ --focus stations


Station S1:
  Utilization: 95% ← Very high
  Avg queue: 6.2 ← Growing queue
  Max queue: 12
  Tasks processed: 850

Station S2:
  Utilization: 62% ← Underutilized
  Avg queue: 1.5
  Tasks processed: 520

Imbalance: S1 handles 62% more tasks than S2

Traffic Bottleneck

Symptoms: 

High wait times
Edge/node occupancy > 80%
Low robot utilization despite tasks available

Verification: 

waremax analyze results/ --focus traffic


Traffic Hotspots:
  Node N15: 92% occupancy
  Node N22: 85% occupancy
  Edge E15: 88% occupancy

Wait Time Analysis:
  Avg traffic wait: 12.3s
  Peak traffic wait: 45s
  Time spent waiting: 28% of task time

Robot Bottleneck

Symptoms: 

Robot utilization > 90%
Tasks queued with no available robots
Idle time = 0

Verification: 

waremax analyze results/ --focus robots


Fleet Analysis:
  All robots: 92-95% utilization
  Idle time: <1%
  Task queue: 15 tasks waiting

This indicates: Not enough robots for demand

Demand Bottleneck

Symptoms: 

Low robot utilization
Low station utilization
Robots often idle

Verification: 

Fleet utilization: 45%
Station utilization: 38%
Throughput: Equal to order arrival rate

This indicates: System capacity exceeds demand

Step 4: Deep Dive Analysis

Time Breakdown

waremax analyze results/ --breakdown task-time

Task Time Breakdown:
  Component       Time    %
  ─────────────────────────────
  Queue wait      3.2s    7%
  Travel time     15.8s   35%
  Traffic wait    10.5s   23%  ← High
  Station queue   8.2s    18%  ← High
  Service time    7.5s    17%
  ─────────────────────────────
  Total           45.2s   100%

Largest delays:
  1. Traffic wait (23%)
  2. Station queue (18%)

Location Analysis

waremax analyze results/ --breakdown location

Congestion by Location:

Top 5 Congested Nodes:
  N15 (intersection):  92% occupancy, 8.5s avg wait
  N22 (station area):  85% occupancy, 6.2s avg wait
  N8 (crossing):       78% occupancy, 4.1s avg wait
  N30 (aisle):         72% occupancy, 3.5s avg wait
  N12 (junction):      68% occupancy, 2.8s avg wait

Step 5: Visualize the Bottleneck

Heat Map View

Warehouse Traffic Heat Map:

   ░░░░░░░░░░░░░░░░
   ░░░░░▒▒▒▒░░░░░░░
   ░░░▒▒▓▓▓▓▒▒░░░░░
   ░░▒▓▓████▓▓▒░░░░  S1 area
   ░░░▒▒▓▓▓▓▒▒░░░░░
   ░░░░░▒▒▒▒░░░░░░░
   ░░░░░░░░░░░░░░░░

   ░ = Low    ▓ = High
   ▒ = Med    █ = Bottleneck

Time Series

Queue Length at S1 Over Time:

12│            ████
  │          ██████████
 8│        ████████████████
  │    ████████████████████████
 4│ ████████████████████████████
  │████████████████████████████████
 0└─────────────────────────────────
   0   15   30   45   60  75  90 min

Queue builds during peak, never clears

Step 6: Confirm with Experiment

Test your hypothesis:

Station Bottleneck Fix

# Increase S1 capacity
waremax run scenario.yaml \
  --param stations.S1.concurrency=3 \
  -o results_fixed/

waremax analyze results/ results_fixed/

Comparison:
                Original    Fixed      Change
Throughput      850/hr      1,020/hr   +20%
S1 utilization  95%         68%        -27%
Avg task time   45.2s       38.1s      -16%

Confirmed: S1 was the primary bottleneck

When bottleneck isn't obvious:

waremax analyze results/ --systematic-bottleneck

Systematic Bottleneck Search:

1. Theoretical max throughput analysis:
   - Station capacity: 1,200/hr
   - Robot capacity: 900/hr ← Limiting
   - Traffic capacity: 1,100/hr

2. Constraint identification:
   Primary: Robot count (fleet too small)
   Secondary: Traffic at N15

3. Utilization analysis:
   - Robots: 95% (saturated)
   - Stations: 72% (headroom)
   - Edges: 65% (headroom)

Conclusion: Add more robots to increase throughput

Step 8: Bottleneck Resolution

Quick Reference

Bottleneck Fix Options
Station Add concurrency, add station, reduce service time
Traffic Widen paths, add routes, improve routing policy
Robots Add robots, improve utilization
Layout Restructure map, move stations

Priority Matrix

Impact    Easy to Fix    Hard to Fix
───────────────────────────────────
High      Fix now        Plan fix
Low       Maybe fix      Don't fix

Common Bottleneck Patterns

The Hidden Bottleneck

Symptom: Low throughput, nothing obviously saturated

Cause: Multiple small bottlenecks compound

Solution: Look at time breakdown, fix biggest first

The Shifting Bottleneck

Symptom: Fix one issue, another appears

Cause: Sequential bottlenecks

Solution: Analyze system holistically, balance capacity

The Interaction Bottleneck

Symptom: Components fine individually, slow together

Cause: Traffic conflicts, resource contention

Solution: Analyze interactions, improve coordination

Next Steps