Elsevier

Computer Communications

Volume 158, 15 May 2020, Pages 39-50
Computer Communications

Resource provisioning towards OPEX optimization in horizontal edge federation

https://doi.org/10.1016/j.comcom.2020.04.009Get rights and content

Abstract

Energy consumption is a key performance metric in multi-access edge computing (MEC) system. Therefore, minimizing consumed energy cost is critically essential. The 5G networks deal with edge computing resources so that the operator would face with power supply limitation. Therefore, it may not be able to provide sufficient resources to ever-increasing user’s requests. One way to compensate such limitation is to form horizontal edge federation (HEF) so that all participant can share the resource capacities as well as request workloads. Energy efficient and ultra-low latency HEF involves the setting of critical factor in each participant: offloading ratios. The decided offloading ratios must provide satisfactory service level to meet latency and physical resource types capacity constraints demanded by requests. Our proposed problem is an energy efficient operational cost (e-OPEX) optimization problem. In this paper, we formulate it as a mixed integer linear program and demonstrate that the problem is NP-hard and proposed a federated multidimensional fractional knapsack based algorithm (FMFK) as our approach. The result shows that the horizontal edge federation based on the FMFK performs better and saves the more e-OPEX as well as serving more input requests compare with the non-federation approach. The experimental results show that our approach save about 40% of e-OPEX specially for high latency sensitive application requests in hotspot zone. It shows around 50% e-OPEX saving in the case of high computation unit cost compare with non-federation approach.

Introduction

The 5G networks coincide with ultra-low latency request and resource provisioning problem for offloaded computation tasks from the ever-increasing user equipment (UE). Although cloud computing resources have been widely used on-demand by distributed users, cloud services are situated far away from UEs and cannot satisfy the offloaded request requirement  [1], [2]. Therefore, it will be an inconsistency with serving ultra-low latency request in low expenditure through tradition cloud-based services [3]. Edge computing can be addressed to serve such kind of requests [4], [5]. In edge computing, the computing infrastructure is deployed on the edge. The computation tasks from UEs can be offloaded to edge nodes which are closer to UEs compared with the tradition cloud data centers. Although edge computing brings more resource closer to UEs, it still requires more active computation resources to meet UEs’ requirement in 5G networks [6].

The power supply limitation in single edge node is the controversial issue. It has been arising a question that how an edge node can provide the high computation resources. Therefore, single edge node structure may not be sufficient to handle the increasing UEs’ request, due to limited computation capacity. On the other hand, offloading the resource request (i.e., computation task) toward a remote cloud data center can cause high backhaul traffic. It may not satisfy latency constraint of the offloaded computation tasks due to propagation delay from the edge nodes to the cloud. This motivates the federation of multiple edge nodes to enable horizontal offloading, i.e., sharing physical resources, and workloads among the edge nodes in a peer-to-peer manner  [16]. In addition, edge operator can offload computation tasks to the edge nodes with more available resources provided that it can meet the service level agreement (SLA) requirements. In edge computing, SLA is an agreement between the edge resource provider and the UEs that guarantees a minimum service is retained. In this research, SLA is based on satisfying the requests-specific latency constraint as well as ensuring the minimum computation and communication resource allocation among the edge nodes. The resulting structure is named horizontal edge federation (HEF). In horizontal edge federation, we set up multiple edge nodes. They can be deployed under different administrative domain. And, they are hosting common applications, which are serving to the UE’s requests. And, the service functionality from various edge nodes are at the same level, although the commutation and communication unit cost can be heterogeneous at each participating edge node.

In this study, we consider non-uniform, and localized user requests toward individual edges nodes. If edge nodes all work alone, each edge node should allocate required capacity through intra-capacity sharing. But, it cannot provide sufficient physical resource capacity due to power supply limitation. So, as to meet the latency constraint specific to each type of request, the resulting total cost can be high. If different edge nodes have different power supply resources and different cost structures (i.e., in terms of the unit costs of computation and communication resources), we may minimize overall cost in HEF by configuring an appropriate setting of resource provision and request distribution on each edge node. Such a configuration should ensure that each edge node can afford the consumed energy-specific operational cost (e-OPEX) for resource allocation. It will meet the latency constraint when handling the local requests, and the remote requests which are offloaded by other edge nodes.

From the perspective of an individual edge node, offloading its own input request loads to other edge nodes in the same HEF (subject to latency constraint and resource capacity) may potentially minimize its own e-OPEX. However, doing so also increases the computation and the communication resource demands, and thus the e-OPEX in other parts of the system. Therefore, the offloading decisions and physical resource provisioning among participating edge nodes interact with one the other. The amount of physical resource capacities should meet the given latency constraint and it is not linear with the amount of imposed request loads. Therefore, allocating the physical resource capacity and setting up the offloading ratios specific to each edge node, in order to minimize overall e-OPEX, turns out to be non-trivial.

In this paper we address the dynamic energy efficient and resource provisioning approach in horizontal edge federation. We propose an optimization problem in order to minimize the total e-OPEX as well as energy efficient offloading ratios specific to edge nodes’ computation resource with regard to the given latency and capacity constraint.

The rest of this paper is organized as follows: Section 2 briefs the related works. Section 3 defines the system model and problem formulation. Section 4 covers solution approach. In Section 5, we present analysis of simulation result and the conclusion of this work is discussed in Section 6.

Section snippets

Related works

Works related to our paper can be divided into two categories: energy efficient computation offloading in edge computing and related environments. And, edge computing energy efficient resource provisioning problem and how our paper differs from them.

Computation offloading is an essential action of edge computing systems. Several aspects affect computation offloading decision such as request type, application-specific resource requirement, edge node’s resource availability, and energy.

System model

We consider a set of m edge nodes EN={e1,e2,,em}, which are controlled and managed under different Multi-access edge computing (MEC) domain. Each edge node ei, 1im can have n type of application instances with different computation capacity and latency requirement. Let Ai={APi1,APi2,,APin} as the set of type of application instances which are provided by edge node ei to requesters, where APik is the instance of application type k, 1kni, that is deployed in edge node ei. Let define sk as

Federated multidimensional fractional knapsack based algorithm

In this section we will propose a knapsack-based solution for energy efficient resource provisioning in HEF. We map our problem as a federated multidimensional fractional knapsack (FMFK) problem where, we consider a knapsack for each application type. The capacity of knapsack is corresponding to total demanded computation resource associated with the related type of application. In order to demonstrate the effectiveness of the proposed technique, we have also implemented the HEF problem using

Evaluation

This section includes some experiments which we have done in order to investigate the performance of FMFK as well as the Gurobi optimal solver for the HEF system. First, we describe the input parameters. Then, we evaluate our model based on the effect of some input parameters. Finally, we present our deduction about what we have achieved.

Conclusion

In this paper, we proposed an energy-efficient resource provisioning problem in horizontal edge federation system, in light of achieving optimum computation task offloading, and the minimum total e-OPEX. We proposed our problem, and modeled it as a mixed integer linear problem. We proved that our proposed problem is mapped into knapsack problem, which is NP-hard. We proposed a federated multidimensional fractional knapsack-based method in order to make decision in terms of intra/inter-resource

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work was partially supported by ministry of Education, Taiwan , under the project: Trusted Intelligent Edge/Fog computing Technology RSC. (project number: 108B568).

References (31)

  • YousefpourA. et al.

    All one needs to know about fog computing and related edge computing paradigms: A complete survey

    J. Syst. Archit.

    (2019)
  • ChenM. et al.

    Reprint of: From cloud-based communications to cognition-based communications: A computing perspective

    Comput. Commun.

    (2018)
  • HsuC. et al.

    Mobile edge computing

    Wirel. Commun. Mob. Comput.

    (2018)
  • LuoJ. et al.

    Ultra-low latency service provision in edge computing

  • LiuJ. et al.

    High-efficiency urban traffic management in context-aware computing and 5G communication

    IEEE Commun. Mag.

    (2017)
  • QiaoY. et al.

    Stochastic resource management for mobile edge computing in 5g networks

  • ZhuQ. et al.

    Task offloading decision in fog computing system

    China Commun.

    (2017)
  • MelendezS. et al.

    Computation offloading decisions for reducing completion time

  • SunH. et al.

    Joint offloading and computation energy efficiency maximization in a mobile edge computing system

    IEEE Trans. Veh. Technol.

    (2019)
  • LiS. et al.

    Energy-aware mobile edge computation offloading for IoT over heterogenous networks

    IEEE Access

    (2019)
  • ChenL. et al.

    Computation peer offloading for energy-constrained mobile edge computing in small-cell networks

    IEEE/ACM Trans. Netw.

    (2018)
  • KimY. et al.

    Mobile computation offloading for application throughput fairness and energy efficiency

    IEEE Trans. Wireless Commun.

    (2019)
  • GuoS. et al.

    Energy-efficient dynamic computation offloading and cooperative task scheduling in mobile cloud computing

    IEEE Trans. Mob. Comput.

    (2019)
  • LiQ. et al.

    Energy-efficient computation offloading and resource allocation in fog computing for internet of everything

    China Commun.

    (2019)
  • PhamQ. et al.

    Decentralized computation offloading and resource allocation for mobile-edge computing: A matching game approach

    IEEE Access

    (2018)
  • Cited by (0)

    View full text