Conference PaperPDF Available

Flexible Multi-Operator RAN Sharing: Experimentation and Validation Using Open Source 4G/5G Prototype

Authors:

Abstract and Figures

In this paper, we present an open source Radio Access Network (RAN) sharing prototype, based on Open Air Interface (OAI) platform, which employs slicing features provided by an open source RAN controller, known as FlexRAN. Accordingly, this paper analyzes the benefits of employing network slicing in the RAN to introduce more flexibility in the configuration of RAN sharing architectures. The contribution of this paper is twofold. Firstly, we propose a flexible RAN sharing architecture where a specific radio slice is allocated to each operator that shares the same RAN, while considering specific Service Level Agreement (SLA) constraints. Secondly, we validate the proposed architecture via a simple radio resource allocation algorithm, which enables dynamic creation and configuration of radio slices, making it possible to transform on-the-fly a “Multi-Operator Core Network” (MOCN) sharing scenario, wherein the spectrum is shared by multiple operators, into a “Multi-Operator RAN” (MORAN) scenario, wherein the spectrum is isolated among the operators that share the same RAN. Emulation results show that the employment of the proposed algorithm enables a more flexible allocation of the radio resources among the sharing operators, providing better performance in terms of end-to-end latency and throughput compared to a static RAN sharing approach.
Content may be subject to copyright.
Flexible Multi-Operator RAN Sharing:
Experimentation and Validation Using Open Source
4G/5G Prototype
Maya Kassis, Salvatore Costanzo, Mohamad Yassin
T´
el´
ecom Sudparis, Orange-Labs, F-92320 Chˆ
atillon, France
Orange-Labs, F-92320 Chˆ
atillon, France
Emails: maya.kassis4@gmail.com, salvatore.costanzo@orange.com, mohamad.yassin@orange.com
Abstract—In this paper, we present an open source Radio Ac-
cess Network (RAN) sharing prototype, based on Open Air Inter-
face (OAI) platform, which employs slicing features provided by
an open source RAN controller, known as FlexRAN. Accordingly,
this paper analyzes the benefits of employing network slicing in
the RAN to introduce more flexibility in the configuration of
RAN sharing architectures.
The contribution of this paper is twofold. Firstly, we propose
a flexible RAN sharing architecture where a specific radio slice
is allocated to each operator that shares the same RAN, while
considering specific Service Level Agreement (SLA) constraints.
Secondly, we validate the proposed architecture via a simple
radio resource allocation algorithm, which enables dynamic
creation and configuration of radio slices, making it possible to
transform on-the-fly a “Multi-Operator Core Network” (MOCN)
sharing scenario, wherein the spectrum is shared by multiple
operators, into a “Multi-Operator RAN” (MORAN) scenario,
wherein the spectrum is isolated among the operators that share
the same RAN. Emulation results show that the employment of
the proposed algorithm enables a more flexible allocation of the
radio resources among the sharing operators, providing better
performance in terms of end-to-end latency and throughput
compared to a static RAN sharing approach.
Keywords:5G, Network Slicing, Open Air Interface, RAN
sharing, FlexRAN controller.
I. INTRODUCTION
The fifth generation of mobile networks (5G) is expected
to host a plethora of services with different Quality of ser-
vice (QoS) requirements. Henceforth, in order to satisfy the
increasing traffic demands, operators need to improve their
network capacity, while maximizing the efficiency of their
network infrastructures. Progressively installing new Radio
Access Network (RAN) nodes could be a potential solution
for satisfying the increasing customers’ demands. However,
the aforementioned solution would face diverse drawbacks in
term of network investment impacting Capital Expenditures
(CAPEX) and Operating Expenditures (OPEX), while leaving
some environmental effects as well. In this context, RAN
sharing is seen as a fundamental and optimal solution for
future mobile network requirements [1], as it permits to share
CAPEX and OPEX expenditures with other partners and
enhance mobile coverage in a cost effective manner [2].
Specifically, RAN sharing is defined as an approach where
the access network equipment, including the antenna, tower,
power, mast and backhaul equipment are shared among mul-
tiple network operators.
Two most commonly RAN sharing use cases are deployed
by network operators around the world. In the first one,
known as Multi Operator Core Network (MOCN), two or more
operators share the same RAN as well as the spectrum. In the
second one, known as Multi Operator RAN (MORAN), all
the RAN equipment (antenna, tower, site, power, etc.) except
the radio carriers are shared among two or more operators [3].
The MORAN approach is the mainly adopted solution as it
guarantees more independence in the sharing process, letting
each operator to be able to control cell-level parameters, and
therefore, enables service differentiation. Indeed, it permits
each operator to use its own portion of spectrum in an
independent manner. However the MOCN approach is more
efficient and less expensive as it enables a better utilization
of the spectrum but this comes at the cost of additional
complexity in handling the sharing process, while it does not
guarantee isolation among the involved operators.
In this paper, we propose a framework that benefits of
MOCN efficiency capabilities, while providing spectrum iso-
lation, that is the main requirement for MORAN scenar-
ios. The essential concept of our prototype is to enable a
dynamic conversion from a MOCN scenario, wherein the
whole spectrum is shared by multiple operators, towards a
MORAN scenario by employing a flexible network slicing
mechanism while presenting the advantages of both cases. In
this direction, network slicing provides logical isolation of the
shared spectrum, while considering satisfaction requirements
of the operators involved in the sharing process [4].
We have implemented the proposed framework in an E2E
mobile network prototype, based on Open Air Interface (OAI)
platform [5]. Note that OAI provides a fully compliant 3GPP-
based open source implementation of the Long Term Evolution
(LTE) eNB protocol stack, while first mature implementations
for 5G scenarios will be soon made available from the open
source community. For the sake of simplicity, we have im-
plemented our framework using a mature 4G release of the
OAI software but it can be easily extended to 5G scenarios.
Moreover our prototype employs a Software Defined Network
(SDN) approach to decouple the data plane from the control
plane [6] of the MAC Layer, by leveraging from the RAN
controller made available by the Mosaic5G project, known
as FlexRAN [7]. FlexRAN abstracts the MAC scheduling
process, moving the control logic, i.e., the implementation
of the scheduling algorithm, into a centralized entity, while
enabling real-time creation and configuration of radio slices.
Using FlexRAN, we have deployed a radio slice for each
operator, providing isolation by associating a different portion
of radio spectrum to each slice.
In order to validate the benefits of introducing the slicing
concept in our RAN sharing use case, we propose a sim-
ple radio resource allocation algorithm that enables dynamic
allocation of spectrum among the operators who share the
same RAN and spectrum, while guarantying isolation and
operators’ Service Level Agreement (SLA) requirements. By
means of emulations, we show that our prototype is capable
of handling the isolation requirements of a MORAN sharing
scenario, while the employment of the proposed algorithm
provides better performance in terms of end-to-end latency
and throughput compared to a static RAN sharing approach.
The remainder of this paper proceeds as follows. In Sec-
tion II, we provide an overview of the state of the art of
RAN sharing and slicing solutions for 4G/5G networks. In
Section III, we present a description of our prototype architec-
ture. In section IV, we introduce the proposed dynamic slicing
algorithm for RAN sharing, while Section V discusses the
performance of the proposed dynamic network slicing solution
for RAN sharing comparing with a static slicing approach.
Finally, Section VI concludes the paper.
II. RE LATE D WO RK
In this section, we present some of the architectures for
RAN sharing proposed in the literature and some related works
on RAN slicing.
In [8], authors presented the concept of a network slice
broker for mobile shared networks to dynamically orchestrate
the resource allocation process between virtual mobile opera-
tors. Such a broker forecasts the resources allocation demands
of the virtual operators, anticipating the reservation of the
demanded resources. In addition, authors in [9] proposed an
enhanced resource allocation coordinator, called hypervisor,
built on top of the physical resources in a shared RAN environ-
ment. The main task of such an hypervisor is to virtualize the
radio resources and to assign them to a different eNB virtual
instance, that represents a different virtual operator, based on
predefined SLAs.
In [10], authors proposed a framework to support the
sharing of 4G cellular networks between multiple operators
by virtualizing the eNBs. This is done by creating logically
independent virtual eNBs on the top of a shared physical
eNB, while enabling resource isolation and coexistence of
independent policies among different virtual eNBs instances.
Moreover, authors in [11] suggested a resource sharing
mechanism for Time Division Duplexing (TDD) networks.
Specifically, authors put forward a framework to evaluate the
potential benefits offered by an application-oriented network
slicing approach and evaluate the performance of high priority
and low priority traffic in a 5G slicing scenario by means of
a simulation campaign.
In the context of RAN slicing, authors in [12] proposed
a solution for enabling an efficient coexistence of IoT and
eMBB slices. The proposed solution allocates more Radio
Resource Blocks (RBs) to the IoT slice before its traffic peaks,
in order to minimize the impact of IoT peak sessions on the
performance of the eMBB services.
In the same context, authors in [13] proposed a dynamic
partition of the physical bandwidth to meet the distinct SLA of
each service, by deploying a slice policy to create customized
network slices. They also used the traffic rule abstraction to as-
sign a portion of the flowspace to a certain slice. This approach
provides fine-grained flow identification and QoS management
policies on 4G and 5G mobile network architecture.
Authors in [14] suggested a fully programmable RAN
sharing architecture to enforce network slices in an OAI-based
testbed that creates multiple Core Networks (CN) instances
with different functionalities. This framework provides an
initial performance evaluation of a scenario with multiple
slices sharing a single RAN and with multiple core networks.
Moreover, an interesting dynamic end-to-end slicing testbed
has been proposed in [15]. In such a testbed, efficient multi-
service RAN slice management and orchestration is enabled
by automatically adding core networks according to the slice
owner’s requirements, while a flexible slicing controller for
virtualized RAN over heterogeneous deployments has been
proposed in [16].
Differently from the architecture proposed in [8], in this
paper we propose an architecture that leverages from SDN by
using a controller on the top of shared eNB. In [9] and [10],
authors proposed two approaches which have not been tested
in real 5G testbeds and do not take into account the constraints
of future 5G architectures. However, our prototype leverages
from SDN technology, while both did not. Nevertheless, in
contrast to what is mentioned in [11], we evaluate the impact
of a dynamic slicing approach compared to a static resource
allocation approach in a RAN sharing architecture using a real
SDN controller and we perform diverse emulation campaigns
to validate the proposed slicing approach in a real 4G/5G
shared RAN prototype. In [12] and [13], authors used a similar
environment to the one adopted by our paper, but the focus was
on intra-operator slicing differentiation requirements while our
focus is on a multi-operator scenario.
Moreover, differently from [14], [15], [16], we mainly focus
on evaluating the impact of slicing for enabling a more flexible
RAN sharing approach from an operator perspective. Our
goal is to evaluate if slicing could be employed to add more
flexibility in MOCN scenarios, while guaranteeing the same
level of isolation provided by MORAN approaches.
III. PROP OS ED RAN S HA RI NG P ROTOTY PE
In this section, we describe the proposed framework imple-
mented over an OAI-based open source platform, which pro-
vides a flexible ecosystem implementation of 3GPP Evolved
Packet Core (EPC) and RAN nodes [5].
Fig. 1: Proposed MORAN Sharing prototype
Figure 1 illustrates an overview of our architecture. It
consists of an OAI eNB connected to two different OAI EPCs,
each representing a different network operator, configured with
a different Mobile Country Code (MCC)/ Mobile Network
Code (MNC) ID. Specifically, the first operator is configured
with MCC/MNC=20895 and is referred in the following
with Slice ID=0, while the second one is configured with
MCC/MNC=20892 and referred in the following with Slice
ID=92. The eNB is connected to a USRP B210 card [17],
which is in turn connected to a Log Periodic antenna oper-
ating at 2.3 GHz. Note that eNB is configured with Time
Division Duplexing (TDD) transmission mode, with 10 MHz
bandwidth, which corresponds to a total number of 50 RBs.
On top of the shared RAN equipment, we have deployed
the FlexRAN controller (Version V2.2), which is a flexible
and programmable platform for software-defined radio access
networks. It acts as a stand-alone centralized control plane that
enables flexible slicing management of the whole spectrum via
a southbound interface, based on Google Protobuf language.
Besides, it offers a set of Northbound APIs that can be used to
monitor the RAN state and to develop radio resource allocation
policies in an abstracted way [7].
We leveraged from the northbound FlexRAN APIs to create
two radio slices, slice ‘0’ which carries the traffic related to
EPC with MCC/MNC=20895 and slice ‘92’ which carries the
traffic related to EPC with MCC/MNC=20892. The prototype
is completed by two commercial smartphones, equipped with
a SIM card programmed according to the aforementioned
MCC/MNC IDs. The two smart-phones are connected to the
shared OAI eNB over the radio interface (LTE Band 40 TDD),
while their traffic is managed by the EPC that belongs to its
MCC/MNC ID.
Note that OAI provides a basic MOCN sharing implementa-
tion solution, wherein the same RAN equipment and spectrum
can be shared among multiple EPCs via an S1-flex interface.
The essential innovation in our proposal is that we make use
of a network slicing mechanism to isolate the same shared
spectrum into two logically isolated slices. Hence, we are
able to move on demand from the baseline MOCN scenario
implemented in OAI to a MORAN scenario and ensure traffic
segregation. Finally, we have implemented a simple algorithm
via a northbound slicing application, built on top of the
FlexRAN controller, to validate our proposed framework and
the effectiveness of the proposed slicing approach.
Note that a video demonstration of our prototype has been
made available at [18].
IV. DYNA MI C SLICING ALGORITHM FOR MORAN
ENVIRONMENT
In this section, we describe the proposed algorithm we
implemented in order to validate our proposed architecture,
with the aim to ensure dynamic reallocation of radio resources
for the two slices that represent the two sharing operators.
Note that the dynamicity in our approach is shown through
redistribution of the amount of RBs in case the traffic starts to
change upwards by borrowing RBs or downwards by donating
RBs from/to other slices.
Note that in the following, the required RBs for each slice
are defined as the amount of RBs being consumed currently
and it is known by calculating the average of the last 5 RBs
bursts used by each slice for the last 5 time slots.
The proposed algorithm can be split in two processes as
follows. The first one, referred to as “Borrow Process”, is
enabled when a slice needs to borrow RBs from other slices or
from a shared room, i.e., a portion of shared spectrum that is
not currently requested by other slices. Conversely, the second
one, referred to as “Donation process”, is triggered when a
slice experiences a low traffic load condition that may let it to
donate an amount of unused RBs to the shared room and/or
to other slices.
Indeed, the common or shared room value indicates the
number of total resources that could be donated by each slice
in a specific time slot. For instance, if slice 0 can donate 4 RBs
and slice 92 can donate 6 RBs, the shared room can be filled
with 10 RBs to be offered to the slices in need. Accordingly,
this value changes periodically based on the slices’ demands
trend.
We propose two thresholds to be agreed and mentioned by
operators in a potential SLA:
A “low” threshold which represents the minimum guaran-
teed number of RBs in the slice (one threshold per slice),
used in the donation process.
An “high” threshold (one per slice), which acts as a
benchmark and is used to trigger the dynamic slicing
algorithm. It is defined as the maximum allowed RBs
of the slice during the current state - 1. Note that this
equation is chosen for implementation needs, as it lets
us to understand that the amount of RBs allocated to a
tagged slice are fully used. For instance, let’s assume
to reserve 30% of the whole bandwidth to a tagged
slice, i.e., 15/50 RBs in case of 10 MHz bandwidth: the
above threshold will be triggered once this slice will use
more than 14 RBs (15 RB -1). Indeed, our testbed will
realize that tagged slice is experiencing high traffic load
conditions.
The operation of transferring RBs between slices is triggered
with respect to the following conditions:
The demanded RBs of a tagged slice is higher than its
high threshold: this means that this slice needs more
resources.
The SLA rules are not broken, i.e., by assuring that
the slice, which donates RBs, will keep its minimum
guaranteed number of RBs as defined by the SLA.
In a nutshell, we have two constraints to be checked: SLA
agreement, explained as thresholds, and the demands of each
slice.
Algorithm 1 Borrow process
1: SETRB=AV G(RB(TTI)),T)
2: SETHigh threshold=RBs in current state 1
3: while RB> H igh threshold do
4: /*This means High Traffic Load: demand of RBs*/
5: if shared room = 0 then
6: trigger donation process = 1
7: else
8: Borrow from the shared room
9: end if
10: end while
In Algorithm 1, we describe the workflow of the borrow
process of our dynamic slicing algorithm, illustrating the
steps where one slice raises the flag of asking for additional
RBs. Firstly, the high threshold is calculated as mentioned
in the second line of both processes. In line 3 the borrowing
procedure is activated if the RBs demanded by a slice is higher
than the high threshold of this slice. If so, this slice is allowed
to borrow RBs from the shared room in case it is not empty,
otherwise, it will trigger the donation process.
Comparatively, the algorithm shown in Algorithm 2 illus-
trates the donation of one slice to the shared room, at the
condition of not breaking the SLA of that slice. Note that the
transfer of RBs from one slice to another one is allowed only
if the donated slice is experiencing low traffic load (lines 6:9).
Moreover, the borrow process will be repeated periodically
regarding all slices. If one slice is in need, it will trigger the
second process, knowing that the amount of RBs to be given
in one burst is always set to maximum one resource block
group, which is 3/50 RBs in our implementation scenario.
Furthermore, revert back to original setup will occur in case
both slices are experiencing high traffic load.
Note that the high threshold value can be used to set up
the priority of each operator, i.e., a high value for the high
threshold decreases the probability of an operator to trigger
the demand for additional RBs from other slices. Conversely,
a low value for the high threshold allows an operator to trigger
such a procedure even with a small amount of traffic load,
giving it higher priority.
It is worth to note that both processes are being called in
one main process which is always running in the background.
In each loop, the main process calculates the average of the
consumed RBs for each slice besides their high thresholds,
and correspondingly triggers either the donation or the borrow
process.
Algorithm 2 Donation process
1: SETRB=AV G(RB(TTI)),T)
2: SETHigh threshold=RBs in current state 1
3: if trigger donation process = 1 then
4: =RBs RBs to be donated
5: if  > low threshold then
6: if RB< H igh threshold then
7: /*This slice is in low load*/
8: Donate an amount of RBs to the shared room
9: trigger donation process = 0
10: else
11: Revert back to SLA initial case
12: end if
13: else
14: DO NOTHING
15: end if
16: end if
V. PERFORMANCE EVALUATION
In this section, we present the results of an emulation
campaign, wherein we evaluated the performance of the pro-
Initial Parameters Value
Percentage of slice 0 10% (5/50 RBs)
Percentage of slice 92 60% (30/50 RBs)
Minimum guaranteed RBs in
slice 0 3
Minimum guaranteed RBs in
slice 92 3
Maximum threshold in RBs of
slice 0 5-1= 4
Maximum threshold in RBs of
slice 92 30-1=29
Trigger for running dynamic
slicing algorithm
maximum allowed
RBs in the current
state - 1
TABLE I: Initial conditions for dynamic slicing scenario
posed dynamic slicing algorithm, as compared to a static
scenario, wherein the spectrum is allocated to each slice in a
static manner. Such emulation campaign was performed using
two Android smart-phones, each one belonging to a different
operator and indeed served by a different Core Network as
described in Section III. Note that in both scenarios, the
slicing approach is applied only in DL direction, which in
our experiments occupies 70% of the whole spectrum.
Moreover, the spectrum is not equally distributed among
the two slices in order to emulate both a worst and best case
scenario wherein one slice suffers from the lack of available
spectrum, while the other one has enough spectrum to serve
its traffic demands. Specifically, in the static scenario, the
slice 0, which is associated to the operator with MNC=95
has been allocated 10% of the whole spectrum, while 60%
of spectrum is reserved for the slice 92. Accordingly, in
the dynamic scenario we aim at showing that the proposed
algorithm is capable of adjusting the aforementioned unfair
spectrum allocation by redistributing the RBs among slices
according to their needs, while respecting the SLA.
Table I summarizes the initial SLA conditions and param-
eters set during the emulation campaign wherein we evaluate
the dynamic scenario that employs the proposed algorithm.
In our evaluation experiments, two metrics were used to as-
sess the performance of the proposed solution: E2E throughput
and latency measured at the end-user side.
Throughput is determined running a “curl” command
through the “termux” APP on the Android smart-phone device,
which allows for collecting E2E performance statistics.
Figure 2 shows the throughput comparison in dynamic
slicing and static slicing mode for slice 0 when downloading a
10 MB file from a server located in its respected EPC, while
downloading different files (of different size) on the second
smart-phone, belonging to the other slice (slice 92).
As it can be seen from Fig 2, the throughput in the static
scenario remains constant no matter the load on the second
slice is. While the throughput of Slice 0 in the dynamic
scenario is much higher for low traffic loads in the slice 92,
Fig. 2: Throughput of slice 0
while it tends to decrease for high loads in the slice 92. The
reason behind this significant gain is that in dynamic mode,
slice 0 was allowed to borrow RBs from slice 92 when it is
needed, with respect of the SLA conditions.
Conversely, Fig 3 shows the comparison of the throughput
observed for the slice 92 in static and dynamic mode when
downloading 10 MB file from the server located in the re-
spected EPC (with MNC=92), while varying the load on slice
0. Even in that case, it can be observed a gain as compared to
the static mode (up to 50 KB/s gain). However, such a gain
is less than for the one achieved by slice 0. This is expected
because slice 92 is configured with better initial conditions,
i.e., it has already at its disposal a sufficient amount of RBs,
and it does not require to borrow too many RBs from the
other slice. However, if possible, the algorithm lets slice 92 to
borrow RBs from other slices, letting it to further improve its
performance compared to the static scenario.
Fig. 3: Throughput of slice 92
The second metric measured in our experiments is the
E2E latency, that is calculated by running a “ping” command
to an “youtube” website on the aforementioned “termux”
application on smart-phone devices, noting that both handsets
were consuming youtube traffic only with the same resolution
at the same time of running the ping command.
Figure 4 shows a comparison of the E2E latency between
Fig. 4: Comparison of E2E Latency - full load on both slices
static and dynamic slicing mode while consuming youtube
traffic on both slices. It can be observed that our algorithm
leads to a significant decrease of latency for slice 0, (from
747 ms to 374 ms) as it was allowed to borrow additional
RBs from slice 92 when needed, with strict respect of the
other slice’s demands. Accordingly, it is worth to note that
the performance of slice 92 is not impacted even if it has
donated RBs to the slice 0.
Indeed, the above results validate the effectiveness of the
proposed framework, i.e., the adoption of a dynamic slicing
approach together with a clear definition of SLA constraints
can be definitely useful for managing MORAN scenarios in
an efficient manner.
VI. CONCLUSION
In this paper, we have proposed a RAN sharing framework
that adopts a network slicing solution to transform on the fly
a static MOCN scenario in dynamic MORAN scenario. The
proposed framework has been implemented in an open source
4G/5G prototype based on OAI, that has been integrated with
the FlexRAN Controller provided by the Mosaic5G project.
We have shown that the proposed framework is capable of
providing the isolation requirements of a MORAN scenario by
segregating the shared spectrum into logically isolated slices,
while associating a different slice to each operator that shares
the same RAN infrastructure.
Moreover, in order to validate the effectiveness of the
proposed framework, we have tested a simple dynamic slicing
algorithm that aims at bringing more flexibility in the resource
allocation process. By means of experimental measurements,
we have shown that the proposed approach enables an effi-
cient redistribution of the radio resources among the sharing
operators, while taking into account multiple factors: traffic
demands, SLA constraints and isolation requirements.
In future works we aim at testing more complex RAN
sharing algorithms and provide a benchmark scheme that will
take into account recent state of the art dynamic schemes.
Moreover, we aim at testing the proposed framework in 5G
Non Stand Alone (5G NSA) scenario first and in a 5G
SA scenario later, when mature open source code will be
made available by the OAI community. Besides, we aim at
testing the proposed framework in large scale environments
with additional user devices, looking also at first open source
solutions which will be made available by the Open RAN (O-
RAN) alliance.
ACKNOWLEDGMENT
This work was partly funded by the European Commission
under the European Union’s Horizon 2020 programme - grant
agreement number 815074 (5G-EVE project). This paper
solely reflects the views of the authors. The Commission is
not responsible for the contents of this paper or any use made
thereof.
REFERENCES
[1] “Nokia Network Sharing White Paper”, available online at
https://www.nokia.com/networks/solutions/network-sharing/, accessed on
Jan. 2020.
[2] “CSG White paper on RAN SHARING MONETIZA-
TION”, available online at https://www.csgi.com/wp-
content/uploads250 RAN Sharing Monetization.pdf, accessed on
Dec. 2019.
[3] Parallel wireless, “Multi-Tenant and Sharing (MORAN and MOCN) avail-
able online at https://www.parallelwireless.com/products/multi-tenant-
and-sharing/, accessed on Dec. 2019.
[4] P.Rost, et al.,“Network Slicing to Enable Scalability and Flexibility in 5G
Mobile Networks, IEEE Communications Magazine, vol. 55, no. 5, pp.
72-79, May 2017.
[5] N. Nikaien, et al., “OpenAirInterface: A flexible platform for 5G re-
search”, ACM Sigcomm Computer Communication Review, vol.44, no.5,
Oct. 2014.
[6] ONF TR-521 Specification, “SDN Architecture”, Issue 1.1, Feb.2016.
[7] X. Foukas, N. Nikaein, M. M. Kassem, M. K. Marina and K. Kontovasilis.
FlexRAN: A Flexible and Programmable Platform for Software-Defined
Radio Access Networks”, ACM CoNEXT, California, USA, Dec. 2016.
[8] K. Samdanis, X. Costa-P´
erez and V. Sciancalepore, “From network
sharing to multi-tenancy: The 5G network slice broker”, IEEE Com-
munications Magazine, Jul. 2017.
[9] L. Zhao, M. Li, Y. Zaki, A. Timm-Giel and C. Gorg, “LTE mobile network
virtualization, Mobile Networks and Applications”, Mobile Networks and
Applications, Vol.16, No.4, Aug. 2011.
[10] J.S. Panchal, R.D. Yates, M.M. Buddhikot, “Mobile Network Resource
Sharing Options: Performance Comparisons”, IEEE Trans. Wireless
Comm., vol.12, no.9, Sep. 2013.
[11] S. Costanzo, R. Shrivastava, K. Samdanis, D. Xenakis, X. Costa-P´
erez
and D. Grace, “Service-oriented resource virtualization for evolving TDD
networks towards 5G”, IEEE Wireless Communications and Networking
Conference (WCNC), Doha, Apr. 2016.
[12] S. Costanzo, I. Fajjari, N. Aitsaadi and R. Langar, “Dynamic Network
Slicing for 5G IoT and eMBB services: A New Design with Prototype and
Implementation Results,I EEE Cloudification of the Internet of Things
(CIoT), Paris, France, July 2018.
[13] E. Coronado and R. Riggio, “Flow-Based Network Slicing: Mapping the
Future Mobile Radio Access Networks, 28th International Conference on
Computer Communication and Networks (ICCCN), Valencia, Spain, July
2019.
[14] A. Ksentini and N. Nikaein, “Toward Enforcing Network Slicing on
RAN: Flexibility and Resources Abstraction”, IEEE Communications
Magazine, vol. 55, no. 6, pp. 102-108, Jun. 2017.
[15] R. Schmidt and N. Nikaein, “Demo: Efficient Multi-Service RAN Slice
Management and Orchestration”, IEEE/IFIP Network Operations and
Management Symposium, Budapest, Hungary, Apr.2020.
[16] R. Schmidt, C. Chang and N. Nikaein, “FlexVRAN: A Flexible Con-
troller for Virtualized RAN Over Heterogeneous Deployments”, IEEE
International Conference on Communications (ICC), Shanghai, China,
July 2019.
[17] “USRP B200/B210 Specification Sheet”, available online at
https://www.ettus.com/product/details/UB200-KIT
[18] DEMO of Proposed RAN sharing framework, available at
https://youtu.be/obwzZ2fIsPs
... Kassis et al. [40] provide a different approach. The authors propose a scheme based on the borrowing and donation of resources. ...
... From the presented RAN scheduling algorithms, NVS [12] and the work by Kassis el al. [40] do not support different types of traffic. While this is supported in the works by Ksentini et al. [13,15], they give the resource distribution using a strict per-scheduling-opportunity reservation. ...
Article
Full-text available
Enforcing network slice isolation in 5G Radio Access Networks (RANs) can degrade a network slicing solution’s ability to efficiently support different types of traffic. In addition, competing virtual network operators may share one network infrastructure that forces network slicing solutions to provide fair treatment between them. This work presents Flexible Priority Scheduling (FPS). This RAN network slicing solution provides isolated treatment to slices with different traffic types, further exploring their needs and flexibility to achieve a more efficient solution. It does so by defining a contract interface that allows a flexible representation of different types of traffic, from heavy throughput services to low latency and bursty traffic. Furthermore, the presented contract representation allows a fair treatment of the tenant’s traffic in the Medium Access Control (MAC) scheduler that enforces this slicing solution: Priority Adaptation Slice Scheduler (PASS).
... In mobile networks, the radio channel is obviously a precious resource, whence the large work on Radio Access Network (RAN) sharing [7]- [9], which is orthogonal to our study: we focus on allocation of resources upstream of RAN: CPU in network nodes and bandwidth in the physical links between them. Much work also exists on virtual network embedding, i.e., deciding where to place virtual nodes and links onto a physical network [3]- [6], [10]. ...
Conference Paper
Full-text available
Within network slicing, Virtual Network Embedding has been vastly studied, i.e., deciding in which physical nodes and links to place virtual functions and links. However, the performance of slices does not only depend on where virtual functions and links are placed, but also on how much resources they can use, which has been mostly neglected in the literature. We thus propose a method for optimal resource dimensioning, via dimensioning capacities of multiple Jackson networks (one per slice) co-existing in the same resource-constrained network. Despite the long history of Jackson networks, we are the first, to the best of our knowledge, to model such a problem. The objective is to minimize energy consumption while satisfying the latency requirements of heterogeneous service providers. We show numerically that our solution is able to achieve both goals, differently from classic approaches, which assume that the amount of resources assigned to slices is fixed a-priori.
Article
Full-text available
We argue for network slicing as an efficient solution that addresses the diverse requirements of 5G mobile networks, thus providing the necessary flexibility and scalability associated with future network implementations. We elaborate on the challenges that emerge when we design 5G networks based on network slicing. We focus on the architectural aspects associated with the coexistence of dedicated as well as shared slices in the network. In particular, we analyze the realization options of a flexible radio access network with focus on network slicing and their impact on the design of 5G mobile networks. In addition to the technical study, this paper provides an investigation of the revenue potential of network slicing, where the applications that originate from such concept and the profit capabilities from the network operator’s perspective are put forward.
Article
Full-text available
The ever-increasing traffic demand is pushing network operators to find new cost-efficient solutions towards the deployment of future 5G mobile networks. The network sharing paradigm was explored in the past and partially deployed. Nowadays, advanced mobile network multi-tenancy approaches are increasingly gaining momentum paving the way towards further decreasing Capital Expenditures and Operational Expenditures (CAPEX/OPEX) costs, while enabling new business opportunities. This paper provides an overview of the 3GPP standard evolution from network sharing principles, mechanisms and architectures to future on-demand multi-tenant systems. In particular, it introduces the concept of the 5G Network Slice Broker in 5G systems, which enables mobile virtual network operators, over-the-top providers and industry vertical market players to request and lease resources from infrastructure providers dynamically via signaling means. Finally, it reviews the latest standardization efforts considering remaining open issues for enabling advanced network slicing solutions taking into account the allocation of virtualized network functions based on ETSI NFV, the introduction of shared network functions and flexible service chaining.
Conference Paper
Full-text available
The provision of service-oriented network resource virtualization, commonly known as network slicing, is envisioned as an answer to the increasing diversity of application demands in evolving 5G mobile networks. Network slicing is helpful in isolating a specified amount of resources in order to accommodate diverse services having heterogeneous requirements that may be conflicting. In this paper, we propose a service-oriented network resource slicing scheme for a Time Division Duplex (TDD) network that, for a pre-defined time duration, forms service specific network slices based on traffic prediction. The aim of the proposed slicing scheme is to enhance the services, Quality of Experience (QoE) and system resource utilization efficiency by introducing a new degree of flexibility upon allocating resources to different tenants. System-level simulation analysis shows that the proposed scheme improves the performance of high priority services and boosts resource utilization at negligible performance loss for low-priority traffic.
Article
Knowing the variety of services and applications to be supported in the upcoming 5G systems, the current "one size fits all" network architecture is no more efficient. Indeed, each 5G service may have different needs in terms of latency, bandwidth, and reliability, which cannot be sustained by the same physical network infrastructure. In this context, network virtualization represents a viable way to provide a network slice tailored to each service. Several 5G initiatives (from industry and academia) have been pushing for solutions to enable network slicing in mobile networks, mainly based on SDN, NFV, and cloud computing as key enablers. The proposed architectures focus principally on the process of instantiating and deploying network slices, while ignoring how they are enforced in the mobile network. While several techniques of slicing the network infrastructure exist, slicing the RAN is still challenging. In this article, we propose a new framework to enforce network slices, featuring radio resources abstraction. The proposed framework is complementary to the ongoing solutions of network slicing, and fully compliant with the 3GPP vision. Indeed, our contributions are twofold: a fully programmable network slicing architecture based on the 3GPP DCN and a flexible RAN (i.e., programmable RAN) to enforce network slicing; a two-level MAC scheduler to abstract and share the physical resources among slices. Finally, a proof of concept on RAN slicing has been developed on top of OAI to derive key performance results, focusing on the flexibility and dynamicity of the proposed architecture to share the RAN resources among slices.
Conference Paper
Although the radio access network (RAN) part of mobile networks offers a significant opportunity for benefiting from the use of SDN ideas, this opportunity is largely untapped due to the lack of a software-defined RAN (SD-RAN) platform. We fill this void with FlexRAN, a flexible and programmable SD-RAN platform that separates the RAN control and data planes through a new, custom-tailored southbound API. Aided by virtualized control functions and control delegation features, FlexRAN provides a flexible control plane designed with support for real-time RAN control applications, flexibility to realize various degrees of coordination among RAN infrastructure entities, and programmability to adapt control over time and easier evolution to the future following SDN/NFV principles. We implement FlexRAN as an extension to a modified version of the OpenAirInterface LTE platform, with evaluation results indicating the feasibility of using FlexRAN under the stringent time constraints posed by the RAN. To demonstrate the effectiveness of FlexRAN as an SD-RAN platform and highlight its applicability for a diverse set of use cases, we present three network services deployed over FlexRAN focusing on interference management, mobile edge computing and RAN sharing.
Article
Resource sharing among mobile network operators is a promising way to tackle growing data demand by increasing capacity and reducing costs of network infrastructure deployment and operation. In this work, we evaluate sharing options that range from simple approaches that are feasible in the near-term on traditional infrastructure to complex methods that require specialized/virtualized infrastructure. We build a simulation testbed supporting two geographically overlapped 4G LTE macro cellular networks and model the sharing architecture/process between the network operators. We compare Capacity Sharing (CS) and Spectrum Sharing (SS) on traditional infrastructure and Virtualized Spectrum Sharing (VSS) and Virtualized PRB Sharing (VPS) on virtualized infrastructure under light, moderate and heavy user loading scenarios in collocated and noncollocated E-UTRAN deployment topologies. We also study these sharing options in conservative and aggressive sharing participation modes. Based on simulation results, we conclude that CS, a generalization of traditional roaming, is the best performing and simplest option, SS is least effective and that VSS and VPS perform better than spectrum sharing with added complexity.