[论文解读] Edge Cloud Offloading Algorithms: Issues, Methods, and Perspectives

Jianyu Wang, Jianli Pan+4 more|arXiv (Cornell University)|Jun 16, 2018

IoT and Edge/Fog Computing参考文献 37被引用 41

一句话总结

本调查分析边缘云卸载算法，提出多主体模型和分类体系，并对单服务器/多服务器策略、在线/离线平衡，以及分区/粒度考虑进行综述。

ABSTRACT

Mobile devices supporting the "Internet of Things" (IoT), often have limited capabilities in computation, battery energy, and storage space, especially to support resource-intensive applications involving virtual reality (VR), augmented reality (AR), multimedia delivery and artificial intelligence (AI), which could require broad bandwidth, low response latency and large computational power. Edge cloud or edge computing is an emerging topic and technology that can tackle the deficiency of the currently centralized-only cloud computing model and move the computation and storage resource closer to the devices in support of the above-mentioned applications. To make this happen, efficient coordination mechanisms and "offloading" algorithms are needed to allow the mobile devices and the edge cloud to work together smoothly. In this survey paper, we investigate the key issues, methods, and various state-of-the-art efforts related to the offloading problem. We adopt a new characterizing model to study the whole process of offloading from mobile devices to the edge cloud. Through comprehensive discussions, we aim to draw an overall "big picture" on the existing efforts and research directions. Our study also indicates that the offloading algorithms in edge cloud have demonstrated profound potentials for future technology and application development.

研究动机与目标

• Characterize the end-to-end offloading process from mobile devices to edge clouds using a three-agent model (mobile devices, communication links, edge servers).
• Classify existing offloading algorithms along five dimensions: destination, load balancing, mobility, partitioning, and partition granularity.
• Summarize representative single-server and multi-server offloading approaches and their optimization formulations.
• Highlight online vs offline load balancing strategies and their applicability under varying resource constraints.

提出的方法

• Introduce a novel taxonomy for offloading consisting of three agents and five classification dimensions.
• Review and summarize MAUI and CloneCloud as single-server offloading approaches with graph-based and VM-based partitioning.
• Discuss ThinkAir and Cloudlet as multi-server/offloading frameworks with dynamic resource provisioning and ad-hoc cloudlets.
• Categorize online and offline balancing approaches, including admission control, primal-dual, stochastic, and load migration methods.

实验结果

研究问题

• RQ1 What are the primary architectural and optimization dimensions that govern edge cloud offloading?
• RQ2 How do single-server and multi-server offloading approaches differ in partitioning, synchronization, and latency/energy trade-offs?
• RQ3 What online and offline load balancing strategies are proposed to handle resource variation and connectivity in edge clouds?
• RQ4 How do partition granularity and scheduling impact performance and energy efficiency in edge-offloading systems?

主要发现

• MAUI reduces smartphone energy and improves latency by leveraging near-edge servers (up to 27% energy savings for video games, 45% for chess, >85% for face recognition) and 4.8x latency improvement.
• CloneCloud enables fine-grained offloading with automatic migration and achieves up to 20x speedup and 20-fold energy reduction in experiments.
• ThinkAir introduces dynamic, parallel VM-based execution across edge servers to improve scalability and latency tolerance.
• Cloudlet architecture enables flexible, component-level offloading with ad-hoc cloud resources at the network edge to boost performance, albeit with synchronization challenges.
• Online balancing approaches model admission control and resource-aware decisions to maximize throughput in edge cloud environments.
• Offline balancing methods (e.g., RIAL) optimize VM migrations and resource usage via multi-criteria decision-making to reduce migration costs and balance load.