tutorial1

Tutorial 1: VM Scheduling and Performance Evaluation

Overview

This tutorial demonstrates the performance evaluation of different VM allocation policies in edge computing environments. The focus is on comparing various VM scheduling algorithms and their impact on system performance when mobile devices offload computational tasks to edge servers.

Scenario Description

Simulation Scenario

Figure 1: Tutorial 1 simulation scenario architecture.

The simulation environment is designed with the following characteristics:

• Mobile Device Constraints: Mobile devices can only offload tasks to edge servers connected to their serving access point (AP)
• Edge Infrastructure: Edge servers operate a variable number of Virtual Machines (VMs) to handle computational workloads
• Localized Processing: Tasks are processed locally at the edge tier, reducing latency compared to cloud-based processing
• Resource Management: Different VM provisioning algorithms are evaluated to optimize resource allocation

Key Features

• Geographically Distributed Edge Servers: Multiple edge servers strategically placed across the network
• Dynamic VM Allocation: Variable number of VMs per edge server based on demand and allocation policy
• Access Point Association: Strict binding between mobile devices and their serving access points
• Performance Metrics Collection: Comprehensive evaluation of different scheduling approaches

VM Scheduling Algorithms

Figure 2: VM scheduling algorithms comparison.

This tutorial implements and compares five distinct VM scheduling algorithms:

1. Random (RND)

• Strategy: Randomly selects an available VM from the pool
• Characteristics:
  • No optimization criteria
  • Provides baseline performance metrics
  • Useful for comparison against optimized algorithms

2. First-Fit (FF)

• Strategy: Selects the first available VM that can accommodate the task
• Characteristics:
  • Simple and fast allocation
  • Minimizes search overhead
  • May lead to suboptimal resource utilization

3. Next-Fit (NF)

• Strategy: Visits hosts in sequential order and selects the first suitable VM found
• Characteristics:
  • Maintains ordering across allocation decisions
  • Balances simplicity with systematic allocation
  • Provides more predictable allocation patterns than random

4. Best-Fit (BF)

• Strategy: Selects the VM with the highest current CPU utilization that can still accommodate the task
• Characteristics:
  • Maximizes resource consolidation
  • Minimizes resource fragmentation
  • May increase task queuing times on heavily loaded VMs

5. Worst-Fit (WF)

• Strategy: Selects the VM with the lowest current CPU utilization
• Characteristics:
  • Distributes load across available resources
  • Minimizes individual VM overload
  • May lead to resource underutilization

---

Note: This tutorial provides fundamental insights into task offloading decision making. Real-world deployments should consider additional factors such as network conditions, task dependencies, security requirements, and user mobility patterns.