Comparative Analysis of Energy Consumption Reduction in Device-Only, Edge-Only, and MCSA Approaches for Inference Tasks

In edge computing, resources such as computing power and memory are limited, making it crucial to allocate them efficiently for various tasks. This article proposes two algorithms to address this challenge: (1) MLi-GD, which calculates the optimal model segmentation, resource allocation, and strategy selection under user mobility, and (2) CBR for resource renting cost, which aims to maximize benefits by leveraging the same resource renting cost. The algorithms are evaluated through experiments, showing their effectiveness in allocating resources efficiently.

Key Points

1. Edge computing faces limitations in resources such as computing power and memory.
2. Two algorithms are proposed to address this challenge: MLi-GD and CBR for resource renting cost.
3. MLi-GD calculates the optimal model segmentation, resource allocation, and strategy selection under user mobility.
4. CBR for resource renting cost aims to maximize benefits by leveraging the same resource renting cost.
5. The algorithms are evaluated through experiments, showing their effectiveness in allocating resources efficiently.

Let’s break it down further

1. Edge computing is a technology that enables computing tasks closer to the source of data, reducing latency and improving performance. However, these edge devices often have limited computing power and memory, making it crucial to allocate resources efficiently.
2. The two proposed algorithms are designed to address this challenge by optimizing the allocation of resources for various tasks in edge computing. MLi-GD calculates the optimal model segmentation, resource allocation, and strategy selection under user mobility, while CBR for resource renting cost aims to maximize benefits by leveraging the same resource renting cost.
3. The MLi-GD algorithm works by introducing a variable 𝑅 into the objective function to represent the alter-native strategies. This allows the algorithm to calculate the optimal model segmentation, resource allocation, and strategy selection under user mobility.
4. On the other hand, CBR for resource renting cost is designed to maximize benefits by leveraging the same resource renting cost. This means that if two tasks have the same resource renting cost, they will be given priority over tasks with higher costs.
5. The algorithms are evaluated through experiments, showing their effectiveness in allocating resources efficiently. By using these algorithms, edge computing can make better use of limited resources, improving performance and reducing latency.
   In conclusion, this article proposes two algorithms to optimize the allocation of resources for various tasks in edge computing. These algorithms are shown to be effective through experiments, demonstrating their potential to improve the efficiency of edge computing systems. By leveraging the same resource renting cost, these algorithms can make better use of limited resources, resulting in improved performance and reduced latency.