Performance degradation and simulation lag when using very large network files (≈1GB+) and dynamically adding vehicles

[akashmi]

I am facing significant performance issues when running SUMO with a large network file (~1GB net.xml). In the future, I plan to scale this setup to ~10GB network files, so understanding the correct and recommended approach is important.

With the current ~1GB network, SUMO shows noticeable lag and jitter during simulation, especially when dynamically adding and routing vehicles via TraCI.

---

Setup

• Network file size: ~1GB (net.xml)
• Planned network size: up to ~10GB
• Vehicle count: ~100, (goal is to add 5000+ vehicles)
• Vehicle creation: Dynamically via TraCI
• Routing: Dynamic route computation
• SUMO mode: GUI and non-GUI (issue observed in both; GUI slightly worse)
• Hardware: Sufficient RAM/CPU (not hitting memory limits)

---

Issues Observed

1. Simulation lag / jitter

   • Vehicles do not move smoothly
   • Simulation struggles to run in real time

2. Slow route computation

   • Route finding takes noticeable time
   • Adding routes via TraCI causes delays

3. Vehicle insertion overhead

   • Adding vehicles in batches causes simulation stalls
   • After insertion, overall simulation smoothness degrades

4. Scaling concerns

   • Issues already visible at ~1GB network size
   • Unclear how to scale to ~10GB networks efficiently

---

Expected Behavior

• Reasonably smooth simulation with large networks (within practical limits)
• Clear guidance on:
  • Recommended configuration for large networks
  • Best practices for routing and vehicle insertion at scale
  • Known limitations or expected trade-offs

[namdre]

the most likely reason is slow routing speed (grows ~ linear in network size).

to verify: run sumo with option --verbose and check reported time spent with routing tasks

recommendation: try offline routing or use a faster routing algorithm: https://sumo.dlr.de/docs/Simulation/Routing.html