Efficient Computing Resource Sharing for Mobile Edge-Cloud Computing Networks

Edge computing refers to computational processes being performed at or near the source of data generation, rather than relying on a centralized data-processing warehouse. This approach minimizes latency because the data doesn’t have to traverse over a network to a data centre; it can be processed locally on devices like smartphones, IoT devices, or local edge servers. Edge computing is particularly beneficial for real-time applications and services that require rapid processing.

Introduction to Mobile Edge-Cloud Computing

• Mobile Applications Evolution: Highlighting the significance and growth of mobile applications like VR, AR, and e-Health.
• Computing Challenges: Increasing computing demands of mobile applications and the limitations of mobile device hardware.
• Edge and Cloud Computing: Both edge and cloud computing provide services to mobile devices, with the cloud handling bulk tasks and the edge offering closer, efficient computing.

Edge-Cloud Cooperative Networks and Business Model

• Cooperation Necessity: Stresses the need for efficient edge-cloud cooperation for future technology, especially given the rapid growth of mobile application demands.
• Profitability and QoS: Challenges in profitability and Quality of Service for both the MEC and the cloud, and the potential of resource sharing.
• Research Gap and Contributions: Addresses the lack of attention on MEC and cloud cooperation in business models and introduces contributions in proposing efficient resource sharing frameworks and solving computing resource management problems.

State of Art Systems: System Design — MEC System Design and Architectures

Let’s start with the system architectures and technologies. In MEC systems, a variety of architectures have been explored, each with unique technical enablers. One key innovation is the heuristic link-path formulations, which have revolutionized mobile network design. Additionally, the integration of MEC-based architectures is pivotal for real-time applications like surveillance, achieved through wireless communications. Another interesting development is the use of microwave power transfer, addressing the challenge of energy efficiency in low-complexity devices. Cost-effective strategies in deploying datacenters also play a crucial role, ensuring adherence to service level objectives.

State of Art Systems: System Optimization — Enhancing MEC through Energy-Efficient Designs

Moving to advanced MEC designs, we see the incorporation of UAVs in air-ground integrated networks. These networks enhance communication, caching, and computing capabilities of the edge network. Another noteworthy development is the two-level edge computing architecture, specifically designed for automated driving services. However, a crucial aspect that requires further exploration is the integration of computing resources between MEC and the cloud, a potential area for substantial system performance enhancement.

Mobile Edge-Cloud Computing Network Model and Problem Formulation

• Network Composition: Consists of N1 MEC servers and N2 cloud servers. MEC servers, often integrated with cellular base stations, serve local mobile users, while cloud servers, distributed globally, cater to remote users via core networks.
• Task Distribution and Latency Sensitivity: The cloud handles a larger volume of computing tasks with typically lower latency sensitivity compared to MEC servers.
• Quality of Service (QoS) Requirements: MEC servers are tasked with completing computing jobs within tight deadlines to ensure QoS, whereas the cloud focuses on prompt task completion.
• Resource Sharing and Economic Model: MEC servers and the cloud are interconnected and can share computing resources. This includes a wholesale strategy where MEC servers can sell excess computing resources to the cloud, and a buyback scheme allowing MEC servers to purchase resources from the cloud during high demand periods, balancing QoS and profitability.

The architecture for mobile edge-cloud computing networks

Source — From research paper

Operation of Mobile Edge-Cloud Computing Networks

• Efficient Framework Design: Introduces a structured framework for managing computing resource sharing between MEC servers and the cloud.
• Time Scales in System Operation: Differentiates between two time scales — ‘time slot’ (tens of minutes) and ‘time interval’ (hundreds of milliseconds).
• Resource Sharing Mechanism: In each time slot, MEC servers wholesale computing resources to the cloud, which remain fixed for that duration. During each time interval, MEC servers can buy back resources from the cloud as needed.
• Dynamic Resource Management: The cloud utilizes resources wholesaled by MEC during a time slot and ensures availability for MEC buybacks during time intervals.
• Operational Process: Outlines the process for determining wholesale and buyback computing resources, including price setting by the cloud and resource allocation adjustments by MEC servers.

Operation Model of MEC Server

• Task arrivals at MEC servers follow a Poisson distribution; workload follows an exponential distribution.
• MEC servers process tasks under a first-come, first-serve policy.
• Computation delay depends on arrival time, workload, unprocessed tasks, and available resources.
• The model ensures computation delay does not exceed a specified upper bound to maintain QoS.
• MEC servers manage wholesaled and buyback resources to uphold QoS constraints.

Operation Model of the Cloud

• Computing Resource Composition: Includes resources from cloud servers and those acquired from MEC servers through wholesale/buyback schemes.
• Resource Allocation: The total available resources at the cloud are a sum of its own resources and the net resources acquired from MEC servers.
• Task Management: Computing tasks at the cloud follow a Poisson distribution, with workload distributions impacting resource requirements.
• Quality of Service (QoS): Focuses on average computation delay as a QoS metric, with an upper bound to ensure efficiency.
• Resource Utilization: Efficient utilization of available resources is crucial to reduce computation delays, thereby enhancing QoS.

Profit Model of MEC and the Cloud

Looking at the profit model, we see that for MEC servers, profit comprises operation costs, income from processing tasks, and resource trading with the cloud. The cloud’s profit model is similar, with additional consideration for the quality of service penalty. This model underscores the financial motivations behind resource management strategies in both MEC servers and the cloud.

Problem Formulation in Mobile Edge-Cloud Computing Networks

• Goals and Challenges: MEC servers aim to maximize profit while ensuring QoS, and the cloud focuses on balancing operation costs and QoS penalties.
• MEC Server’s Trade-Off: Balances wholesale income against buyback costs, with constraints on wholesaled resources, QoS, and buyback resource limits.
• Cloud’s Optimization: Focuses on trading computing resources with MEC servers, affected by wholesale pricing, and managing local operation costs for optimal QoS.
• Joint Optimization: The interdependence of MEC and cloud resources through wholesale and buyback schemes requires coordinated management to maximize overall profits.

Wholesale and Buyback Scheme without Profit Transfers

• The scenario where MEC servers and the cloud are owned by the same entity, eliminating profit transfers.
• Absence of profit transfer between MEC servers and the cloud.
• Minimize local operation costs and Quality of Service (QoS) penalties at the cloud.

Optimal Resource Allocation

• Equal distribution of computing resources across both MEC servers and the cloud.
• both fairness in resource allocation and operational efficiency.
• Enhanced network management and resource utilization.

Computation Delay and Task Scheduling

Estimating Computation Delay

• Focus on calculating the average time delays for tasks in both MEC servers and the cloud.
• Essential for efficient network management and planning.

Computation Delay and Task Scheduling

Task Scheduling Strategies

First-Come-First-Served (FCFS) Policy:

• Applied when task delays are within acceptable limits.
• Ensures orderly and fair task processing.

High Priority Processing:

• Triggered when task delays exceed set thresholds.
• Tasks are processed faster at alternate MEC servers or the cloud.

Efficiency Boost:

• Task offloading and priority handling significantly enhance network efficiency.

Wholesale and Buyback Scheme with Profit Transfers

Separate Ownership of MEC and Cloud

• In this model, MEC servers and the cloud are managed by different entities or departments.
• This setup introduces the concept of profit transfers between the two.

Dual Objectives for Profit Maximization

• Both MEC servers and the cloud focus on maximizing their profits.
• They aim to achieve this through shared resources and reducing operational costs.

Interlinked Challenges in Resource and Pricing

• Resource allocation and pricing decisions are closely linked due to the need for profit transfers.
• The interconnection of these two aspects adds complexity to management decisions.

Optimal Pricing and Resource Management

Direct Link Between Resources and Pricing

• More resources sold, the higher the wholesale price.
• Prices for resources don’t decrease as more. are sold

Evaluation Setting

• 40 MEC (Mobile Edge-Cloud) servers with 3.2GHz each.
• 10 cloud servers, max 200GHz computing resources.
• Task arrival: MEC servers [6–12] tasks/sec, Cloud 500 tasks/sec.

Task and Service Parameters

• Average computing workload per task: 100Kb.
• Computing resource requirement: 2000 cycles/bit.
• Service fees at MEC and cloud: $0.2764/Gb.
• Buyback resource costs: $0.2736/(GHzhour) and $0.8208/(GHzhour)².
• Task deadline at MEC: 2 seconds, at Cloud: 4 seconds.
• QoS penalty: $104 times delay.

Comparison of Schemes

• Compared with: “No sharing” and “Constant price” schemes.
• No sharing: MEC cannot share resources, buys from cloud for QoS.
• Constant price: Fixed wholesale price at $0.2487/(GHz*hour).

Simulation Results: Algorithm Performance

• Solution found within 15 iterations.
• Parameters increase with more computing tasks at MEC.
• Reasons: MEC reserves more resources, cloud adjusts price and resources for QoS.

Optimal Cloud Resources and Profits

• “No sharing” uses most cloud resources.
• Social welfare maximization uses least cloud resources.
• Total profits increase with higher arrival rates at MEC.
• Social welfare maximization yields highest total and cloud profit but lowest MEC profit.

Wholesale and Buyback Resources

• MEC servers wholesale more resources for profit with increasing workloads.
• Higher arrival rate at MEC means fewer wholesaled resources.
• Buyback resources small compared to wholesaled due to high buyback cost.

System Parameter Effects on Optimal Decisions

• Cloud computing resources under social welfare always smaller than under optimal pricing.
• Increasing cloud workload increases demand for resources.
• Higher computing workload at MEC increases cloud resources and wholesale price.

Resources

https://ieeexplore.ieee.org/document/9046758

In conclusion, the exploration of Mobile Edge-Cloud Computing Networks reveals a future where seamless integration and optimized resource sharing are key to enhancing both network performance and profitability. This journey into the intricate world of MEC and cloud computing underscores the importance of innovative approaches in meeting the ever-evolving demands of technology and user experience.