tutorial4

Tutorial 4: Server Capacity Planning for Vehicular Edge Computing

Overview

This tutorial demonstrates the performance evaluation of different capacity planning approaches in vehicular edge computing environments. The focus is on comparing various algorithms that determine optimal computational resource allocation across edge servers to efficiently serve dynamic vehicular workloads on smart highways.

Scenario Description

Vehicular Edge Computing Architecture

Figure 1: Tutorial 4 vehicular edge computing infrastructure.

The simulation environment is designed with the following characteristics:

• Vehicular Network Constraints: Vehicles can only offload computational tasks to edge servers connected to their serving access point
• Heterogeneous Edge Infrastructure: Edge servers run host machines with varying computational capacity
• Dynamic Workload: Vehicle-generated tasks create time-varying computational demands
• Capacity Optimization: Different edge server capacity planning algorithms are evaluated to optimize resource allocation

Smart Highway Environment

Figure 2: Smart highway simulation environment with circular road topology.

The smart highway simulation includes:

• Circular Road Topology: Continuous traffic flow on a closed-loop highway system
• Variable Vehicle Population: 1000 to 2000 vehicles traveling simultaneously
• Dynamic Velocity Modeling: Realistic speed variations based on vehicle position and traffic conditions
• Geographical Distribution: Non-uniform vehicle density creating varying computational demands

Capacity Planning Algorithms

This tutorial implements and compares three distinct approaches for edge server capacity planning:

1. Random Capacity Allocation

• Strategy: Randomly distributes total computational capacity across edge servers
• Total System Capacity: 220 GIPS (Giga Instructions Per Second)
• Characteristics:
  • No optimization criteria or traffic awareness
  • Provides baseline performance metrics for comparison
  • Random distribution may lead to resource imbalances
  • Useful for evaluating the impact of intelligent capacity planning

2. Equal Capacity Distribution

• Strategy: Uniformly distributes computational capacity across all edge servers
• Individual Server Capacity: 20 GIPS per host machine
• Total System Capacity: 220 GIPS (11 hosts × 20 GIPS)
• Characteristics:
  • Simple and fair resource allocation approach
  • Ensures consistent computational capability at each location
  • May not adapt to traffic density variations
  • Provides stable performance baseline

3. Traffic Density Heuristic

• Strategy: Distributes capacity proportionally based on traffic intensity patterns
• Adaptive Allocation:
  • High Density Areas: 44 GIPS computing capacity per host
  • Medium Density Areas: 20 GIPS computing capacity per host
  • Low Density Areas: 14 GIPS computing capacity per host
• Total System Capacity: 220 GIPS (dynamically allocated)
• Characteristics:
  • Traffic-aware resource allocation strategy
  • Maximizes computational resources where demand is highest
  • Adapts to real-time vehicular density patterns
  • Optimizes overall system efficiency

---

Note: This tutorial provides fundamental insights into capacity planning for vehicular edge computing. Real-world deployments should consider additional factors such as vehicle mobility prediction, emergency service prioritization, network handover optimization, and energy efficiency constraints.