5G-COMPLETE: End-to-end Resource Allocation in Highly Heterogeneous Beyond 5G Networks

Full text

5G-COMPLETE: End-to-end Resource Allocation
in Highly Heterogeneous Beyond 5G Networks
Agapi Mesodiakaki, Marios Gatzianas, George Kalfas
Department of Informatics, Aristotle University of Thessaloniki
Center for Interdisciplinary Research and Innovation, Thessaloniki
Francesca Moscatelli, Giada Landi,
Nicola Ciulli, Leonardo Lossi
Nextworks, Pisa, Italy
Abstract—5G-COMPLETE project deals with the delivery of
novel technological blocks in terms of both physical connectivity
and network softwarization for 5G and beyond 5G (B5G) net-
works. The solutions developed within the project employ service-
driven slice management while integrating the orchestration of
radio, transport and core network resources; including both
communication and computational ones (e.g., Multi-Access Edge
Computing (MEC) and cloud computing). In this paper, we
focus on the end-to-end (E2E) resource optimization framework
proposed within 5G-COMPLETE, while highlighting its ability
to perform resource allocation targeting at network energy
efficiency optimization, and consequently, decreased operational
expenditure in 5G and B5G networks. Simulation results are also
provided to demonstrate the significant gains of the developed
resource allocation algorithms compared to the state-of-the-art.
Index Terms—Multi-Access Edge Computing, Energy Effi-
ciency, Management and Orchestration, Beyond 5G Networks,
End-to-end Network Slicing, Network Optimization.
I. INTRODUCTION
Beyond 5G (B5G) networks are expected to meet a plethora
of service requirements supporting high resource and tech-
nology heterogeneity. At the same time, they are expected to
include a variety of different technologies of distinct char-
acteristics and requirements. To this end, 5G-COMPLETE
[1] aims to deliver a plethora of novel technological blocks
targeting at the physical layer as well as at the softwariza-
tion layer of the network, exploiting: i) the advancements
in analog and digital optics assisted by powerful baseband
processor platforms for 100 Gbps and beyond interfaces, ii)
Software Defined Networking (SDN)-enabled mmWave nodes
for point-to-multipoint (P2MP) edge networking where the
fiber deployment is unavailable, iii) wideband sub-THz radio
frontends for unlimited capacity, and iv) the efficiency of
high-speed and time-sensitive packet-switched transport for
synchronized and accurate networking at the edge. Aside from
hardware innovations in the communication infrastructure, the
project invests on the development of powerful Multi-Access
Edge Computing (MEC) platforms for the rapid and cost-
efficient service deployment through Virtual Machines (VMs),
containers and unikernels.
Focusing on such complex and heterogeneous architectures,
network resource planning becomes challenging. Adopted ar-
chitectural paradigms, such as Centralized-Radio Access Net-
work (C-RAN) and Network Function Virtualization (NFV),
have introduced new challenges (e.g., Virtual Network Func-
tion (VNF) placement [2]) to traditional user association and
traffic routing problems.
First, each requested service, which can be viewed as an
ordered set of VNFs (e.g., firewall, NAT, etc.), referred to
as Service Function Chain (SFC), should be deployed at the
network so that not only the correct VNF ordering is ensured
(i.e., proper SFC chaining) but also the capacity constraints
of the host nodes and links are satisfied [3]. The NFV-
supporting physical infrastructure contains a variety of nodes,
including traffic-forwarding switches and computational nodes
of different capabilities able to host VNFs in the form of VMs,
containers or unikernels. Among them, there are cloud servers,
located at distant locations, offering high computational power
at low cost, and MEC nodes able to bring computational
capabilities very close to the user (often being attached to Base
Stations (BSs)), while offering ultra-low latency at the expense
of high cost. Second, apart from fiber links interconnecting
the physical machines, wireless links may also be deployed
as an X-haul transport solution, to offer high flexibility to the
Radio Access Network (RAN). Millimeter wave (mmWave)
constitutes a very promising candidate to serve this purpose,
due to its high bandwidth availability and antenna gains that
are able to compensate for the higher path loss in this band.
In such a complex setup, network resource planning should
jointly consider: i) all different types of technologies, e.g., 5G-
NR, mmWave, fiber, along with their benefits and constraints,
and ii) the allocation of all different resource types, i.e., com-
munication, computational and storage. It is also crucial to
consider the whole network path from the traffic source to the
destined User Equipment (UE), so as to satisfy the service
latency constraint and offer true End-to-End (E2E) optimality.
Furthermore, achieving high energy efficiency is of utmost
importance not only to limit the network operator’s operational
costs (thus, increasing its revenue) but also to decrease the
ICT carbon footprint, leading to eco-friendly B5G networks.
Consequently, energy-efficient network resource planning so-
lutions are needed that will jointly solve the user association,
VNF placement (SFC chaining) and traffic routing problem in
highly heterogeneous B5G networks (HetNets).
A. Related work and Contribution
The joint problem of VNF placement, SFC chaining and
routing has been widely studied in wired networks, mostly
targeting cloud environments in the network core [3], [4],

[5], [6], [7]. However, most of these works do not con-
sider mobile networks and, even when they do, they ignore
the wireless Access Network (AN) segment. Models with a
distinct MEC/cellular flavor appear in [8], [9]; however, [8]
does not consider power consumption, while [9] lacks the use
of specific and detailed link constraints, while targeting at a
different objective, i.e., minimizing maximum link utilization.
To this end, in this work, we advance the state-of-the-art
(SoA) by proposing a novel resource management framework,
developed within 5G-COMPLETE, for E2E network perfor-
mance optimization incorporating the model and algorithms
of [10]. We focus on the joint VNF placement and routing in
a MEC/cloud-enabled mobile HetNet consisting of macro BSs
and SCs. We explicitly include the AN, as well as wireless X-
haul links, while, unlike [11], adding all necessary controls for
the computational resources, the associated delay and capacity
constraints. The developed algorithms are compared to refer-
ence solutions, while their interaction with the Management
and Orchestration (MANO) framework is being detailed. The
proposed framework can be employed by a mobile network
operator as an offline tool, during the network planning stage,
to provide quantitative answers on the power expenditure and
computational resources (both of them major components of
OPEX/CAPEX) required to support a given set of services.
The remainder of the paper is structured as follows. Sec-
tion II presents the 5G-COMPLETE resource allocation frame-
work targeting at E2E network performance optimization.
Section III analyzes the MANO framework focusing on the
monitoring platform. In Section IV, the main results of the
developed algorithms are presented and discussed compared
to the SoA. Finally, Section V draws the main conclusions.
II. 5G-COMPLETE E2E RESOURCE OPTIMIZATION
In this section, we present the employed system model and
we discuss the studied problem, while analyzing the main
aspects of its formulation (a complete description together
with detailed formulas can be found in [10]). Due to the high
complexity of the derived optimal solution, in this section, we
also propose a low-complexity algorithm, which targets at high
energy-efficiency, while ensuring high UE acceptance ratio.
A. System Model
We consider a network consisting of a set of nodes (exclud-
ing mobile users) and a set of links among them, as illustrated
in Fig. 1. The set of nodes refers to gNBs and/or SCs (hereafter
referred to as BSs) in the RAN, as well as switches/routers and
other middlebox devices (e.g., load balancers, firewalls, etc.)
in the Core. These devices typically operate as VNFs running
in virtual instances (i.e., VMs, containers and unikernels) uti-
lizing computational resources collocated with network nodes.
The considered links can be divided into the X-haul links, i.e.,
fronthaul, midhaul, backhaul, that can be wired or wireless
(mainly mmWave) and access network links, between the
UEs and their serving BSs. Each link is characterized by a
communication capacity and a delay (the sum of transmission
and propagation delays) to all packets traversing it. We also
: Middleboxes (VN F-
implemented)
: Fiber link
: gNB, SC : MEC resources
: Fog resources
: Switch
Aggreg.
Layer 1
Aggreg.
layer 2
: Cloud resource s : X-haul
wireless link
: Access link
Fig. 1. Mobile network of heterogeneous communication and computation
nodes along with middlebox functionality offered by deployed VNFs. Dashed
color lines indicate 2 illustrative E2E paths selected for 2 highlighted UEs.
consider a set of UEs, each one connecting to a single BS.
We focus on Downlink (DL) and assume that each BS has
a maximum number of Physical Resource Blocks (PRBs) to
allocate to its associated UEs. A subset of network nodes have
also computational (i.e., CPU1) resources able to host one or
more VNFs, whose number is limited by the maximum amount
of CPU resources (measured in GFLOPS) available at each
node.
We also consider a set of SFCs, i.e., ordered sequences of
VNFs. The traffic of each SFC is properly served only if it
passes through all its VNFs exactly matching the specified
order. Each VNF is viewed as a complete software stack that
can be deployed in the computational nodes. Please note that
we allow for multiple instances of a given VNF in the same
or different nodes depending on network traffic. Each VNF
is described by four identifiers referring to: a) the VNF’s
functionality (e.g., NAT etc.), b) the data processing capacity
of the VNF (in Mbps), c) the amount of CPU resources (in
GFLOPS) required for the VNF’s operation, and d) the data
processing delay experienced by an individual data packet as
it passes through the VNF.
Each UE requests a service type based on its SFC, to
be transmitted from the “source” to the “destination”, i.e.,
to the UE, while satisfying a specific E2E throughput and
maximum allowed latency. We explicitly allow for sharing a
VNF instance among two (or more) different SFCs. Also, for
each UE, we use our knowledge (or estimates) of the SINR
received by each BS and its requested service rate to compute
the number of required PRBs. Assuming frequency-flat slow
fading [11], we consider uniform BS power allocation among
the PRBs, so that each PRB is assigned a specific amount of
power (different between gNBs and SCs).
B. Studied problem
Targeting at high energy-efficiency, we employ resources,
devices and links only when needed. Specifically, collocated
CPU resources at a specific node are employed only when
1Additional computational resources such as memory and storage can be
similarly handled and are omitted for simplicity and without loss of generality.

the node actually runs deployed VNFs. Hence, the power
consumed by CPU processing at a specific node is given by
the summation of the idle power of the CPU (when the CPU
is active) and a variable part that depends on the node’s load
[6]. The same applies to communication power consumption
in fiber links. For the switches, the total power is equal again
to the sum of the idle power of the switch and a variable term,
which, in this case, depends on the number of active ports, i.e,
the active fiber links of the switch [6]. The power expenditure
of the mmWave links is proportional to the number of RF
chains of the link and, again, contains an idle part (when the
link is active) and a load-dependent one [11]. The latter applies
also to the power consumption of the AN links (both gNBs
and SCs) [11].
Other constraints include: a) the load of BSs, BH links,
and CPUs can not be higher than their maximum respective
capacity, b) each requested VNF within an SFC must be
deployed in the exact order as defined by the required SFC,
c) each UE must properly attach to exactly one BS, d) the
number of deployed VNF instances on a node is sufficient to
meet the data processing requirements for the incoming traffic,
while ensuring that the computational node is deployed only
when needed, e) the flow conservation and routing conditions
are met, and f) the E2E delay constraints of the UEs are
satisfied. Due to the high complexity of exactly solving the
aforementioned problem via an Integer Linear formulation
[10], we next propose a heuristic algorithm and use the former
as a benchmark so as to evaluate its performance.
C. Proposed Heuristic (HERO)
Our proposed Heuristic for Energy-efficient VNF place-
ment, traffic Routing and user assOciation (HERO) aims at
maximizing the network energy efficiency while ensuring low
UE blocking probability. HERO consists of two stages, as
shown in Fig. 2. In the first stage, the traffic path, from the
source of the requested SFC to the destination UE, is selected
(i.e., user association and routing), while in the second one
VNF placement takes place, ensuring correct VNF ordering.
Initially, to ensure high UE acceptance ratio, the UEs are
sorted based on their requested E2E latency, giving priority to
the UEs with the most delay-intolerant services. For UEs with
the same delay requirements, priority is given to the UEs with
higher rate demands. Then, for each UE, a weighted graph
is constructed from the service source to the UE with all
feasible links, i.e., links whose capacity constraints are not
violated with the association and routing of the studied UE,
and their respective power consumptions acting as weights.
HERO calculates the kshortest-weighted paths and starts with
the first one, as long as it satisfies the delay and link capacity
constraints, taking into account the network state (i.e., the
available capacity of the BSs and BH links due to the already
associated UEs). Otherwise, the next path is selected until
either a feasible path is found or there are no other paths. In
the latter case, the UE is being blocked and HERO proceeds
with the next UE, as long as there is one.
Input
For each UE:
Construct a weighted graph including:
Feasible AN links, X-haul transport links, and Fiber links
Edge weights : Link power c onsumption taking into acc ount any network change
Calculate the
k
shortest-weighted paths
First stage
Output
Move to the next
path
Satisfies
delay &
capacity
constraint s?
Other
VNFs?
YES NO
Put UE to unsati sfied user s
Define UE examination or der
YES
NO
N
Other
paths
available?
User association &
traffic routing
We select the (next) node with the highest
H
to place the
VNF ensuring correct VNF ordering
YES
YES
NO
Change
selected path
For each VN F of the UE we sort the c omputing nodes
of the path bas ed on parameter
H
that considers:
Node centrality
Computational power
Utilization
YES
Other nodes
available in
the path?
NO
Comp.
capacity
constraint
satisfied?
Place the
VNF
Second stage
VNF
placement
hth
s
ers
Y
E
S
p
y
c
?
constraint s?
c
?
NO
NO
availabl
e?
f
the UE we sort the computing nodes
a
sed on paramet er
H
t
h
at
co
n
s
i
de
r
s
:
N
ode centralit
y
Computational power
Select the (next) shortest-weighted path
Ed
Other
UEs?
NO
Othe
r
VNFs
?
P
ut UE
Move
to
the
next
path
Y
E
We select the
(
next
)
V
NF
ensu
Y
E
S
S
F
C

U
C
omp.
capa
city
co
n
st
r
ai
n
t
sati
sfie
d?
s
?
Place the
VNF
to unsatis
f
ie
d us
F
or each VNF
o
of
the path
ba

N

C
p
Sele
c
UEs
?
N
O
YES
NO
VNF
hest
H
to
place
orderin
g
YES
YES
YES
NO NO
e
the
e
the
NO
Update the network
TUBUF
Fig. 2. Flowchart of the proposed energy-efficient VNF placement, traffic
routing and user association algorithm (HERO).
After a valid path is found for the current UE, HERO
proceeds to the second stage, where the VNFs comprising
the UE’s SFC are being placed. For each VNF, following the
order of the user’s SFC, a list is constructed with all available
computational nodes based on a parameter, denoted by H.
This parameter is equal to the sum of the normalized node
centrality, the normalized node computational capabilities and
the node CPU utilization. The latter is equal to:
a) 1, when the studied VNF can be placed in the examined
node without initiating a new VNF instance,
b) 0.1, when there is enough computational capacity to host
the studied VNF in the examined node, but a new VNF
instance is required,
c) 0, otherwise.
Subsequently, the node with the highest Hfor the selected
VNF is selected, as long as it has sufficient computational
resources to host it. Otherwise, the node with the next highest
His selected, until either the VNF is placed at some node
or there is no other node to examine in the selected path. In
the latter case, the algorithm returns to stage 1 and the next
path out of the kis examined. The process is repeated for the
new path until either all VNFs of the UE are placed or there
is no other path to study and the UE is blocked. In case all
UE VNFs are placed, the network state is updated, and the
algorithm proceeds to the next UE. The aforementioned steps
are repeated until all UEs are studied.
III. NETWORK MANAGEMENT &ORCHESTRATION
In this section, we present the 5G-COMPLETE orchestra-
tion framework, and describe its functionalities and how the

aforementioned resource allocation algorithms are integrated
to it.
A. Service-driven Slice Modelling and Management
The 5G-COMPLETE orchestration framework, whose high-
level architecture is depicted in Fig. 3, implements a central
management and coordination layer that provides mechanisms
for the dynamic provisioning and configuration of Verti-
cal Services in tailored E2E heterogeneous Network Slices,
i.e., including different technology components belonging to
the radio, core and transport network. The Vertical Service
Management Function (VSMF), based on the Vertical Slicer
prototype [12], represents the unified entry point of the
5G-COMPLETE ecosystem and exposes procedures for the
instantiation of Vertical Services, by specifying high-level
requirements related to the service business logic. At this layer,
the goal is to hide the actual network complexity, enabling at
the same time cross-technology deployments in a transparent
manner. Indeed, the Vertical Service high-level requirements
are automatically mapped into a set of 5G QoS Indicators
[13] within a Network Slice Type (NEST) template [14],
which represents the Network Operator’s offer for the 5G
System deployment and configuration of the requested Service.
The instantiation request is then propagated to the Network
Slice Management Function (NSMF), where the NEST is
associated with a 3GPP-based [15] Network Slice Template
(NST) that formalizes 5G requirements related to the mobile
connectivity and describes the service characteristics through
the composition of technology-specific Network Slice Subnets
(NSSs). Indeed, the E2E NST, which results from the compo-
sition and sharing procedure performed at the NSMF, includes
three technology-specific Network Slice Subnet Templates that
target respectively the RAN, the 5G Core, and the Vertical
application. In particular, when the latter three components are
virtualized and dynamically orchestrated, the corresponding
subnet refers to the NFV Network Service Descriptors (NSDs)
for the actual deployment and configuration of the VNFs
composing the service. The latter is performed through the
interaction with an NFV/MEC Orchestrator at the underlying
layer and according to the specified SFC.
B. Monitoring Platform
The 5G-COMPLETE Monitoring Platform, depicted in
Fig. 4, is designed to support the collection, normalization
and aggregation of heterogeneous data from different kinds of
data sources and at different layers of the 5G-COMPLETE
architecture. Indeed, software and infrastructure components
in the 5G-COMPLETE ecosystem can export metrics relevant
to the optimization of resources and services. The Monitoring
Platform also provides a configuration service that allows us to
initiate specific monitoring jobs, e.g., related to the monitoring
of infrastructure and virtualized resources within a target E2E
Network Slice. The proposed monitoring solution is suitable
for the integration of different kinds of algorithms (e.g., online
or AI/ML based), while the collected data can be consumed
through the RESTful API exposed by Prometheus [16], as
Fig. 3. 5G-COMPLETE Multi-layer Orchestration Framework Architecture.
the baseline monitoring engine, or through a topic-specific
subscription model in the provided message bus [17] for a
near real-time consumption of relevant information.
C. Resource Allocation for E2E Network Slicing
The NSMF is the Orchestration framework’s component in
charge of the actual Lifecycle Management (LCM) of the E2E
Network Slice, which is performed through the coordination
of its constituent NSSs and the consequent interaction with
technology specific platforms at the NSS Orchestration layer:
i) WAN Infrastructure Managers (WIMs) for the configuration
of the X-haul Transport Network (TN), ii) technology-specific
SDN controllers targeting specifically the fronthaul, the mid-
haul or the backhaul, iii) NFV/MEC Orchestrators for the
orchestration of VNFs, iv) RAN management platforms (e.g.,
the O-RAN Service Management and Orchestration - SMO
[18]). During the instantiation phase, the NSMF is assisted by
the Placement Service (see Fig. 3), which embeds the resource
allocation algorithms and is capable of determining which
computational and network resources should be allocated
for a given service, by performing the joint computation of
optimal network paths along with the VNF placement. The
Placement Service collects relevant data/metrics to feed the
resource allocation algorithms via a monitoring system (see
Fig. 4). Through the processing of these metrics, the Placement
Service supports on-demand allocation of both network and
computational resources, providing to the NSMF the mapping
between subnets’ virtualized components (i.e., VNFs) and
target host nodes as well as the list of connectivity services
to be established in order to guarantee the proper service
operation. According to the computed optimal allocation,
the NSMF coordinates the E2E Network Slice provision-
ing by requesting the NSSs instantiation and configuration
to specialized Network Slice Subnet Management Functions
(NSSMFs). Each NSSMF triggers the underlying technology-
specific platforms to handle the orchestration of its target NSS,
e.g., both the RAN, the 5G Core and the Vertical application
NSSMFs interact with the NFV/MEC Orchestrator for the
VNFs provisioning in the target host nodes, while the TN
NSSMF requests the provisioning of connectivity services to
the WIM. The latter in turn triggers the technology-specific

Fig. 4. 5G-COMPLETE Monitoring Platform.
SDN controllers for the network paths enforcement across the
front/mid/back-haul network.
IV. PERFORMANCE EVALUATION
In this section, we present the considered simulation sce-
narios and derived results for the performance evaluation of
the studied algorithms. Main insights, derived by the provided
results, are also provided.
A. Simulation scenario
The resource allocation heuristics were developed in MAT-
LAB and the optimal solution was computed by IBM CPLEX.
A gNB sector area of 500 m radius is considered overlaid with
two SC clusters [11]. Each cluster consists of four SCs, which
are uniformly dropped in an 100-m-radius from the cluster
centers. One SC (randomly selected) per cluster and the gNB,
are assumed to be fiber-connected to the aggregation network,
while mmWave X-haul links are deployed among BSs. The
aggregation network consists of two layers each one consisting
of four nodes, fiber-connected among them and with the BSs,
so that there is no disconnected node. Hotspot UE traffic is
also assumed [11].
We consider 5 different SFCs of specific VNF ordering, E2E
delay requirement and share of the total requests, as shown in
Table I. In the same Table, the data processing capacity and
the GFLOPS requirement of each VNF type are presented.
For a given number of UEs, we run 10 different scenarios,
with 5 different UE distribution snapshots each. We assume
100 PRBs allocated per gNB or SC (corresponding to μ=0
in 5G numerology). The gNB and SCs operate at 2 GHz,
assuming orthogonal channels between the gNB and SCs.
However, the SCs of different clusters may interfere with each
other. The mmWave X-haul links operate at 60 GHz, with
200 MHz channel bandwidth. For the AN and the mmWave
links, we employ the link budget equation and the related
parameter values of [11]. The number of cores is equal to 8,
24 and 48 for the MEC, 1st and 2nd Aggregation Layer nodes,
respectively. The maximum CPU power is selected randomly
from the set {55, 70}for the MEC nodes, from {150, 220}
for the 1st Aggregation Layer and from {200, 278}for the
TABLE I
SFC AND VNF DETAILS [6]
Type VNF ordering Throughput Delay Share
(Mbps) (ms) (%)
Web NAT-FW-TM-WOC-IDPS [0.6-1] 500 20
VoIP NAT-FW-TM-FW-NAT [0.404-0.64] 100 20
Streaming NAT-FW-TM-VOC-IDPS [5-24] 100 39
Gaming NAT-FW-VOC-WOC-IDPS [0.24-0.5] 60 6
Ultra RT AI/ML NAT-NAT [15-25] 1 15
Type NAT FW TM VO C WOC IDPS
Process Capacity (Mbps) 500 400 200 578 300 600
GFLOPS Requirement 110 440 55 110 110 440
2nd Aggregation Layer nodes, while the idle power is assumed
to be equal to 10%of it. The fiber link capacity is 10 Gbps,
while the rest of the simulation parameters are selected as in
[10].
The optimal solution and the proposed heuristic (HERO)
are compared with:
a) Holu, a heuristic which first performs the VNF place-
ment based on node centrality and CPU utilization of
computing nodes and then decides upon traffic routing. In
addition, it targets at minimizing the power consumption,
while satisfying the E2E service delay constraint,
b) BCSP, which considers node centrality for the VNF
placement and the shortest-path in terms of delay for
routing, while meeting the E2E service delay constraint.
Given that both algorithms do not handle user association, we
employ the default user association for both, i.e., the UEs are
connected to a gNB or SC based on the highest received signal.
In addition, for a fair comparison, their UE examination order
is selected to be the same with HERO.
B. Simulation results
In Fig. 5 and 6, we show the energy efficiency (bits/Joule)
and the computational time (s, in logarithmic scale), respec-
tively, of all algorithms for different number of UEs. As can
be seen, HERO provides a very good trade-off between energy
efficiency and complexity compared to other approaches,
achieving up to 78%of the optimal value, with up to 742 times
lower complexity. All algorithms have 100%user acceptance
ratio in all cases except for BCSP that presents 96%for N=20,
93%for N=30, and 91%for N=40. This is due to the fact that,
in BCSP, the CPU utilization of the computing nodes is not
taken into account, resulting in less efficient VNF placement,
which under higher traffic load can lead to few UEs being
blocked. Compared to Holu and BCSP, HERO achieves up
to 60%and 86%higher energy efficiency, respectively, while
keeping the complexity low2, as shown in Fig. 6. This is due to
the fact that HERO jointly considers user association leading
to higher flexibility at the expense of a little higher complexity.
On the other hand, in both Holu and BCSP, the serving BSs
are already decided (based on the best SINR criterion) and
2All algorithms have been simulated in similar environments, and thus, their
relative performance difference is representative for their evaluation compari-
son. Their absolute simulation times, however, can change considerably, e.g.,
by applying hardware acceleration techniques.

60%
86%
Fig. 5. Energy efficiency (bits/Joule) of all algorithms for different number
of UEs per gNB area.
then the optimal VNF placement and traffic routing from the
UE traffic source to its serving BS are performed. This further
justifies the motivation of our work that user association, VNF
placement and traffic routing should be jointly considered to
guarantee true optimal E2E network performance.
V. F UTURE WORK &CONCLUSION
In this paper, the 5G-COMPLETE resource management
framework was presented targeting at true E2E network per-
formance optimization. In particular, we studied the joint
VNF placement, user association and traffic routing in B5G
networks targeting at energy efficiency maximization, while
ensuring high UE acceptance ratio. We solved this problem
by modelling it with minimal assumptions, while capturing all
characteristics of the employed technologies, resource and ser-
vice types as well as their constraints and power consumption.
To tackle the prohibitive complexity of the studied problem,
we proposed HERO, an energy-efficient resource planning
heuristic, which was shown to significantly outperform the
SoA, while achieving up to 78%of the optimal value, with
up to 742 times lower complexity. To provide a holistic
view of the applicability of the proposed resource allocation
framework, we also analyzed its integration within the MANO
architecture, focusing on the monitoring platform and the
interaction between the components.
ACKNOWLEDGMENT
This work was supported by the H2020 5G-COMPLETE
project (Grant Agreement No 871900).
REFERENCES
[1] (2021, Feb.) 5G-COMPLETE website. [Online]. Available: www.
5gcomplete.eu
[2] A. Laghrissi and T. Taleb, “A survey on the placement of virtual
resources and Virtual Network Functions,” IEEE Commun. Surveys Tuts.,
vol. 1, no. 2, pp. 1409–1434, 2019.
[3] J. Liu, W. Lu, F. Zhou, P. Lu, and Z. Zhu, “On dynamic Service
Function Chain deployment and readjustment,” IEEE Trans. Netw.
Service Manag., vol. 14, no. 3, pp. 543–553, Jun. 2017.
Fig. 6. Execution time (s) in logarithmic scale of all algorithms for different
number of UEs per gNB area.
[4] Y. Ma, W. Liang, M. Huang, W. Xu, and S. Guo, “Virtual Network
Function service provisioning in MEC via trading off the usages be-
tween computing and communication resources,” IEEE Trans. on Cloud
Comput., Dec. 2020, early access.
[5] S. Ayoubi, S. Sebbah, and C. Assi, “A logic-based Benders decomposi-
tion approach for the VNF assignment problem,” IEEE Trans. on Cloud
Comput., vol. 7, no. 3, pp. 894–906, Dec. 2019.
[6] A. Varasteh, B. Madiwalar, A. van Bemten, W. Kellerer, and C. Mas-
Machuca, “Holu: power-aware and delay-constrained VNF placement
and chaining,” IEEE Trans. Netw. Service Manag., Jan. 2021.
[7] M. Tajiki, S. Salsano, L. Chiaraviglio, M. Shojafar, and B. Akbari, “Joint
energy efficient and QoS-aware path allocation and VNF placement for
Service Function Chaining,” IEEE Trans. Netw. Service Manag., vol. 16,
no. 1, pp. 374–388, Mar. 2019.
[8] D. Dietrich, C. Papagianni, P. Papadimitriou, and J. Baras, “Network
function placement on virtualized cellular cores,” in Proc. 9th Intl. Con-
ference on Communication Systems and Networks (COMSNETS), Jan.
2017.
[9] S. Yang, F. Li, S. Trajanovski, X. Chen, Y. Wang, and X. Fu, “Delay-
aware Virtual Network Function placement and routing in edge clouds,”
IEEE Trans. Mobile Comput., vol. 20, no. 12, pp. 445–459, Feb. 2021.
[10] M. Gatzianas, A. Mesodiakaki, G. Kalfas, and N. Pleros, “Energy-
efficient Joint Computational and Network Resource Planning in Beyond
5G Networks,” in Proc. Globecom, Madrid, Spain, 2021, to appear.
[11] A. Mesodiakaki, E. Zola, R. Santos, and A. Kassler, “Optimal user
association, backhaul routing and switching off in 5G heterogeneous
networks with mesh millimeter wave backhaul links,” Ad Hoc Networks,
vol. 78, pp. 99–114, Sep. 2018.
[12] C. Casetti, C. Chiasserini, J. Mart´
ın-P´
erez, N. Molner, and T. Deiss, “The
vertical slicer: Verticals’ entry point to 5G networks,,” in Proc. EuCNC,
Ljubljana, Slovenia, 2018.
[13] “TS 23.501 Technical Specification Group Services and System Aspects;
System architecture for the 5G System (5GS); Stage 2 (Rel. 16),” 3GPP,
2020, v. 16.6.0.
[14] “Generic network slice template,” GSMA, 2020, v. 4.0.
[15] “TS 28.541 5G Management and orchestration; 5G Network Resource
Model (NRM); Stage 2 and stage 3 (Rel. 16),” 3GPP, 2020, v. 16.7.0.
[16] (2021, Sep.) Prometheus Monitoring System. [Online]. Available:
https://prometheus.io/
[17] (2021, Sep.) Kafka Streaming Platform. [Online]. Available: https:
//kafka.apache.org/
[18] (2021, Sep.) O-RAN Alliance. [Online]. Available: https://www.o-ran.
org/