IEEE Communications Magazine - June 2017 - page 26

IEEE Communications Magazine • June 2017
24
0163-6804/17/$25.00 © 2017 IEEE
A
bstract
In order to cope with the increasing network
utilization driven by new mobile clients, and to
satisfy demand for new network services and
performance guarantees, telecommunication ser-
vice providers are exploiting virtualization over
their network by implementing network services
in virtual machines, decoupled from legacy hard-
ware accelerated appliances. This effort, known
as NFV, reduces OPEX and provides new busi-
ness opportunities. At the same time, next gen-
eration mobile, enterprise, and IoT networks are
introducing the concept of computing capabili-
ties being pushed at the network edge, in close
proximity of the users. However, the heavy foot-
print of today’s NFV platforms prevents them
from operating at the network edge. In this arti-
cle, we identify the opportunities of virtualization
at the network edge and present Glasgow Net-
work Functions (GNF), a container-based NFV
platform that runs and orchestrates lightweight
container VNFs, saving core network utilization
and providing lower latency. Finally, we demon-
strate three useful examples of the platform: IoT
DDoS remediation, on-demand troubleshooting
for telco networks, and supporting roaming of
network functions.
I
ntroduction
Data consumption is growing exponentially in
today’s communication networks. This irrevers-
ible trend is driven by the increase of end users
and the widespread penetration of new mobile
devices (smartphones, wearables, sensors, etc.).
In addition, mobile data consumption is also
accelerated by the increased capabilities of the
mobile clients (e.g., higher resolution screens
and HD cameras) and the user desire for high-
speed, always-on, multimedia-oriented connec-
tivity. In numbers, it has been estimated that
connected devices will exceed 50 billion, gen-
erating zettabytes (1 billion terabytes) of traffic
yearly by 2020.
At the same time, the telecommunication ser-
vice provider (TSP) market is becoming compet-
itive with the rise of many over-the-top service
providers lowering subscription fees for users.
Moreover, today’s TSPs often experience poor
resource utilization, tight coupling with specific
hardware, and lack of flexible control interfaces,
and fail to support diverse mobile applications
and services. As a result, TSPs have started to
lose existing and new revenue, while suffering
increased capital and operational expenditure that
cannot be balanced by increasing subscription
costs [1].
In order to cope with the aforementioned
challenges, service providers have started to
softwarize their network infrastructure. By vir-
tualizing traditional network services (e.g., fire-
walls, caches, proxies, intrusion detectors, WAN
accelerators), providers can save operational
and capital expenses, and satisfy user demands
for customized and rapidly evolving services.
This transformation, referred to as network
functions virtualization (NFV), transforms how
operators architect their network to decouple
network functionality from physical locations
for faster and flexible network service provi-
sioning [1]. NFV has gained significant attention
since its first appearance in 2012, resulting in
many, albeit still preliminary, deployments at
the providers’ data centers.
While NFV is gaining attention, a new, fifth
generation (5G) mobile architecture is being
designed to support the increased user demand
mentioned above [2]. As a key design objec-
tive, 5G mobile networks will utilise mobile (or
multi-access) edge computing (MEC), an IT ser-
vice environment with cloud computing capabil-
ities at the edge of the home, enterprise, or IoT
network, within close proximity to the mobile sub-
scribers [3], as shown in Fig. 1. While there have
been a few proof of concept (PoC) implementa-
tions of MEC (including, e.g., a video streaming
deployment at the Wembley Arena
1
), these PoCs
do not present a generic NFV architecture, nor do
they support today’s customer edge devices (e.g.,
home routers).
In this article, we present Glasgow Network
Functions (GNF), an NFV platform that brings
NFV and edge computing together by using
generic, lightweight Linux containers to host virtu-
al network functions (VNFs) in a distributed, het-
erogeneous edge infrastructure. We present three
useful applications and show that by utilizing the
network edge, for example, home, enterprise, and
Internet of Things (IoT) edge, providers can alle-
viate their unnecessary core network utilization
(which can correspond to savings of millions of
dollars per year), perform better troubleshooting
on their network, and provide location-transpar-
ent services to their users.
Container Network Functions:
Bringing NFV to the Network Edge
Richard Cziva and Dimitrios P. Pezaros
A
dvances
in
N
etworking
S
oftware
The authors identify the
opportunities of virtualiza-
tion at the network edge
and present Glasgow
Network Functions, a con-
tainer-based NFV platform
that runs and orchestrates
lightweight container
vNFs, saving core network
utilization and providing
lower latency. They
demonstrate three useful
examples of the platform:
IoT DDoS remediation,
on-demand troubleshoot-
ing for telco networks,
and supporting roaming
of network functions.
The authors are with the University of Glasgow.
Digital Object Identifier:
10.1109/MCOM.2017.1601039
1
2017)
1...,16,17,18,19,20,21,22,23,24,25 27,28,29,30,31,32,33,34,35,36,...228
Powered by FlippingBook