IEEE Systems, Man and Cybernetics Magazine - January 2023 - 12

operations. These tasks can be allocated in separated
nodes; for example, the firewall function can be allocated in
the mobile nodes. Hence, the whole task can be allocated in
two paths: one for latency-critical jobs and the other for
small jobs. The common placement method will allocate
their VNFs in resource-sufficient nodes such as path A
in Figure 1(a).
In Figure 1(b), we can see that
there are two paths to allocate
VNFs. We assume that the devices
in path A have sufficient resources
to place VNFs at a high cost, while
the nodes in path B have limited
resources at a low cost. Next, we
assume that the task of remote
operating can, respectively, place
different functions in these paths.
In addition, these nodes are all
NFV enabled for allocating VNFs.
The systems choose edge nodes
rather than remote clouds because
the remote clouds cannot meet the
latency requirement of this critical
task. As we mentioned before, the cost of placing VNF
blocks in nodes 1 and node 2 is higher than in nodes 3 and
4. To meet the resource constraints, the video transmission
job has to allocate sufficient nodes. If the operating
Node 1
Path A
Path A
Node 2
cost of node 1 is US$100, then the cost of node 2 is 120.
Both nodes have sufficient resources to place their VNFs;
we will choose the lower cost in our example. The job of
instruction transmission requires fewer resources with
strict latency requirements. Therefore, we can allocate this
function in low-cost nodes 3 and 4 and then choose an
appropriate node on the basis of the latency requirement.
Furthermore, when processing
Numerical simulation
shows that our
proposed method
can significantly cut
down the cost while
meeting the multiple
resource constraints.
costs vary from node to node and
the communication costs vary
from link to link, the physical
nodes and links with lower costs
should be considered. Therefore,
this idea reduces the costs and
increases the profits of both providers
and users. The objective of
this article is to minimize the cost
and profit as the supplementary
objective in our model.
Related Works
We review prior works on the VNF
placement problem in two aspects:
Source
Source
Node 3
Node 1
Path A
(a)
Path A
Node 4
Node 1
Destinatio
Destination
cloud and edge. Most works are discussed at the cloud
level. As we know, NFV brings a novel view of operating
network functions. Gao et al. [3] introduce an optimal
framework of the VNF to meet diverse objectives; VNF
problems are considered as service chains for a series of
functions that implement the required functions. Zheng et
al. [4] propose a multiple resource allocation for VNF
deployments, such as bandwidth, central processing unit
(CPU), and random-access memory (RAM). The reliability
of VNF placement is considered in [5]; the authors propose
a reliability-aware resource allocation method with
sharing redundancy in the service chain, and the cost of
redundancy is also considered.
VNFs and their replica placement (VNRFP) are a revoSource
Source
Path
B
Node 3
(b)
NFV-Enabled Nodes
Switches
Figure 1. An illustration of the motivation of the VNF
placement that is cost-efficient for various edge
nodes. (a) Scenario a. (b) Scenario b.
12 IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE January 2023
Path B
Node 4
Destinatio
Destination
lutionary solution that Abdelaal et al. [6] offer for the VNF
placement challenge (VNFRP). While taking into account
various resource limits, their solution attempts to achieve
load balancing over the core links. Accordingly, the VNF
placement problem is initially described as an integer linear
programming (ILP) optimization problem with the goal
of reducing the cost of Service Function Chain (SFC)
installation and the link bandwidth, energy, and consumption
costs. Zeydan et al. [7] employ multiattribute decisionmaking
algorithms for VNF placement inside the cluster to
improve the VNF service satisfaction rates and reduce
deployment costs. Cost is also a main concern in the placement
problem; Gao et al. [8] present a VNF placement and
scheduling method in the cloud, while this article focuses
on the industry edge scenario.
From the perspective of the edge, most research works
focus on latency and dynamic requirements. By combining
the benefits of edge computing and cloud computing, satellite
edge clouds can make it easier for many IoT users to
IEEE Systems, Man and Cybernetics Magazine - January 2023

Table of Contents for the Digital Edition of IEEE Systems, Man and Cybernetics Magazine - January 2023
