[Paper Review] ENORM: A Framework For Edge NOde Resource Management

Nan Wang, Blesson Varghese+2 more|arXiv (Cornell University)|Sep 12, 2017

IoT and Edge/Fog Computing16 references4 citations

TL;DR

ENORM is a lightweight framework for managing edge node resources in fog computing, enabling dynamic provisioning and auto-scaling of workloads on resource-constrained edge devices like routers and base stations. Evaluated on a Pokémon Go-like game, ENORM reduced application latency by 20%-80% and cut data transfer and cloud communication by up to 95% compared to cloud-only deployment.

ABSTRACT

Current computing techniques using the cloud as a centralised server will become untenable as billions of devices get connected to the Internet. This raises the need for fog computing, which leverages computing at the edge of the network on nodes, such as routers, base stations and switches, along with the cloud. However, to realise fog computing the challenge of managing edge nodes will need to be addressed. This paper is motivated to address the resource management challenge. We develop the first framework to manage edge nodes, namely the Edge NOde Resource Management (ENORM) framework. Mechanisms for provisioning and auto-scaling edge node resources are proposed. The feasibility of the framework is demonstrated on a PokeMon Go-like online game use-case. The benefits of using ENORM are observed by reduced application latency between 20% - 80% and reduced data transfer and communication frequency between the edge node and the cloud by up to 95\%. These results highlight the potential of fog computing for improving the quality of service and experience.

Motivation & Objective

• To address the lack of resource management frameworks for edge nodes in fog computing architectures.
• To enable efficient provisioning and deployment of workloads on resource-constrained edge nodes, such as routers and base stations.
• To develop an auto-scaling mechanism that dynamically adjusts CPU and memory resources for edge-hosted applications without compromising primary services.
• To validate the feasibility and performance benefits of fog computing using a real-world, latency-sensitive application use-case.
• To reduce dependency on cloud servers by offloading computation and communication to edge nodes, improving QoS and reducing network overhead.

Proposed method

• Proposes a lightweight provisioning mechanism using a linear search algorithm to initialize services on edge nodes, minimizing overhead.
• Introduces a dynamic auto-scaling mechanism that adjusts CPU and memory resources in response to workload demands.
• Uses containers instead of VMs for workload deployment, due to lower resource overhead on edge devices.
• Employs a handshaking protocol between cloud and edge nodes to manage service lifecycle, including deployment and termination.
• Applies a unit-based scaling approach where resources are increased or decreased in fixed increments to balance responsiveness and over-provisioning.
• Designs the framework for multi-tenant edge environments, ensuring primary network services (e.g., routing) are not disrupted.

Experimental results

Research questions

• RQ1 How can edge nodes be provisioned and integrated into a cloud-based application ecosystem with minimal overhead?
• RQ2 What is an effective and lightweight method for deploying and managing workloads on resource-constrained edge nodes?
• RQ3 How can edge resources be dynamically scaled to meet service-level objectives without affecting the primary function of the edge node?
• RQ4 To what extent can fog computing reduce application latency and cloud communication frequency in real-time, location-aware applications?
• RQ5 Can a lightweight, container-based auto-scaling mechanism achieve significant QoS improvements in edge deployments?

Key findings

• ENORM reduced application latency by 20% to 80% compared to a cloud-only deployment in a Pokémon Go-like online game use-case.
• Data transfer and communication frequency between edge nodes and the cloud were reduced by up to 95% due to localized processing.
• The framework demonstrated feasibility in real-world conditions with low computational overhead, confirming the viability of edge-based resource management.
• The auto-scaling mechanism effectively balanced resource usage and performance, improving QoS without over-provisioning.
• The use of containers instead of VMs enabled efficient workload deployment on edge nodes with limited hardware resources.
• The linear search-based provisioning and scaling mechanisms proved effective and lightweight, suitable for constrained edge environments.