Resource provisioning towards OPEX optimization in horizontal edge federation
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 edge nodes , which are controlled and managed under different Multi-access edge computing (MEC) domain. Each edge node , can have type of application instances with different computation capacity and latency requirement. Let as the set of type of application instances which are provided by edge node to requesters, where is the instance of application type , , that is deployed in edge node . Let define 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)
- et al.
All one needs to know about fog computing and related edge computing paradigms: A complete survey
J. Syst. Archit.
(2019) - et al.
Reprint of: From cloud-based communications to cognition-based communications: A computing perspective
Comput. Commun.
(2018) - et al.
Mobile edge computing
Wirel. Commun. Mob. Comput.
(2018) - et al.
Ultra-low latency service provision in edge computing
- et al.
High-efficiency urban traffic management in context-aware computing and 5G communication
IEEE Commun. Mag.
(2017) - et al.
Stochastic resource management for mobile edge computing in 5g networks
- et al.
Task offloading decision in fog computing system
China Commun.
(2017) - et al.
Computation offloading decisions for reducing completion time
- et al.
Joint offloading and computation energy efficiency maximization in a mobile edge computing system
IEEE Trans. Veh. Technol.
(2019) - et al.
Energy-aware mobile edge computation offloading for IoT over heterogenous networks
IEEE Access
(2019)