Edits.md

SDN Smart Routing – Environment & Code Fix Report

1. Controller Used

This project uses the POX SDN controller:

https://github.com/MurphyMc/pox

Run the controller with:

python3 pox.py openflow.discovery forwarding.l2_learning SDN-Smart-Routing.Smart_Routing_Module

Structure:

pox/
├── pox.py
├── sdn_smart_routing/
│   ├── __init__.py
│   └── Smart_Routing_Module.py
└── ...

---

2. Python Environment Requirements

⚠️ This project is based on legacy SDN libraries and is not fully compatible with Python 3.12 without version constraints. (Ubuntu 24 LTS)

---

Required Packages

Create a requirements.txt:

networkx>=2.0,<3.0
fnss==0.9.1
pyzmq
numpy>=1.21,<2.0
scipy>=1.7,<2.0
matplotlib
python-igraph
fastnumbers
mininet

---

Installation Note (Tested Stable Combo)

pip install "fnss==0.9.1" "networkx>=2.0,<3.0"

and

apt install mininet

---

3. Key Code Fixes

---

3.1 Edge Normalization (CRITICAL)

The graph is undirected, but dictionary keys depend on ordering.

Problem

(3, 2) != (2, 3)

Solution

Always normalize edges:

edge = tuple(sorted(Edges[i]))

---

Helper Function (Recommended)

def edge_key(e):
    return tuple(sorted(e))

Usage:

edge = edge_key(Edges[i])

---

3.2 Locations Requiring Fix

Edge normalization must be applied in:

• CC computation loop

• MTBF computation loop

• MTTR computation loop

• Any access to:

  Links_Lengths_Dictionary[Edges[i]]

---

3.3 Correct Implementations

CC computation

edge = tuple(sorted(Edges[i]))
cc.append(Links_Lengths_Dictionary[edge] / minimum)

---

MTBF computation

edge = tuple(sorted(Edges[i]))
MTBF.append((cc[i] * 365 * 24) / Links_Lengths_Dictionary[edge])

---

MTTR computation

edge = tuple(sorted(Edges[i]))
mttr = round(Links_Lengths_Dictionary[edge] * Gama[i])

if mttr < 1:
    mttr = 1

MTTR.append(mttr)

---

3.4 Link Initialization Fix

edge = tuple(sorted(Edges[i]))

L.append(
    Links(
        Edges[i],
        i,
        Links_Lengths_Dictionary[edge],
        MTBF[i],
        MTTR[i],
        0, 0, 0, True
    )
)

---

3.5 Stochastic Variable Fixes (CRITICAL)

Problem

numpy arrays used instead of scalars

---

TTF Fix

TTF = np.random.exponential(scale=L[i].MTBF)
L[i].Next_Failure = round(TTF) + 1

---

Moderator Fix

moderator = np.random.uniform(0.1, 0.9)

---

Log-Normal Recovery Time

Log_Normal = np.random.lognormal(mu, sig)

print(
    'The link', L[link].ID,
    'will wait up to', round(Log_Normal),
    'to get recovery'
)

---

Second Failure Time (TTF2)

TTF2 = np.random.exponential(scale=L[link_return].MTBF)

---

4. Root Cause Summary

Issues were caused by:

• Undirected graph edge ordering inconsistency
• Mixing numpy arrays and scalar values
• Deprecated SciPy API usage
• FNSS + NetworkX version mismatch
• Incompatibility with Python 3.12 ecosystem

---

5. Final Recommendation

For stable execution:

• Use Python ≤ 3.10 (recommended: 3.7)
• Normalize all edges using tuple(sorted(edge))
• Ensure all stochastic outputs are scalars
• Avoid numpy arrays in scheduling logic
• Use compatible versions of NetworkX and FNSS

---