You are on page 1of 5

Carrying MTC Service in 5G - A Network

Management Perspective
(Invited Paper)
Xu Li, Jaya Rao, Hang Zhang, and Sophie Vrzic
Huawei Technologies Canada Inc., Ottawa, Ontario CANADA
Email: {xu.lica, jaya.rao, hang.zhang, sophie.vrzic}@huawei.com

Abstract—MTC services are undergoing rapid increase; billions instructions to data plane hardware for optimizing the opera-
of MTC connections can be anticipated in a few years. When tion of the entire network. TE is to jointly determine for each
supporting MTC services in telecommunications networks, MTC traffic flow the communication paths (i.e. routing) and resource
traffic characteristics and its unique service requirements should allocation along the paths (i.e. traffic splitting), with respect
be exploited in order to optimize resource utilization and offer to individual flows’ QoS requirements, and network resource
satisfactory service quality for all devices. In this paper, we
constraints so that a network utility is maximized [6]. Flows
present an Information-Centric Virtual Network Architecture
(IC-VNA) that allows to meticulously combine both the customer are split among their routing paths in the data plane, following
and operator specific virtual functions along with their require- control plane instructions.
ments to construct an augmented virtual network topology. It is While SDN provides the framework for achieving central-
shown that through the IC-VNA it is feasible to rapidly roll-out ized control of the network and as a consequence, poten-
highly programmable and flexible services while providing high tially globally optimal network performance, another degree
degree of scalability and cost effectiveness for both MTC service of freedom for facilitating MTC services within the context of
providers and network operators. 5G networks is realized via the concept of network function
virtualization (NFV) [7]. In this regard, NFV which is a
technology complementary to the SDN, allows implementing
I. I NTRODUCTION various network functions such as deep packet inspection (DPI)
Machine type communications (MTC) is a technology that and network address translation (NAT) as virtual software
realizes a system of networks for collecting data from ma- entities on commercial off-the-shelf (COTS) hardware. This
chines such as sensors and smart meters which are often capability effectively decouples the network functions from
massively and densely deployed [1]. MTC applications are re- the physical equipment. The SDN controller itself can be
sponsible for translating received raw machine data into mean- instantiated as a virtualized network function in any high-
ingful information to support decision making, e.g. earthquake capacity processor residing within either a data center, server
prediction, or automation, e.g. defense mechanism activation. or a network node. From the context of MTC, it is imperative
Unlike current world-wide human-centric networks, MTC net- for the network to support multitudes of services, each possibly
works are featured with absence of direct human intervention distinguished by a sequence of application specific functions.
and rapid increase in connections. It is anticipated that MTC As such, the combined SDN and NFV technologies provide
will expand to 2.1 billion connections by the year 2021 [2]. a compelling solution for facilitating the rapid roll-out of
When supporting MTC services, network resources are shared highly programmable and flexible services while providing
between machines and hand-held devices. In order to optimize high degree of scalability and cost effectiveness for both MTC
resource utilization and deliver the best quality of service service providers and network operators.
(QoS) to all devices and users, MTC traffic characteristics and In the sequel, we give an overview of MTC services, share
varied service requirements should be taken into account when our solution to MTC services in 5G, which leverages SDN
performing network management and traffic control. and NFV, and discuss relevant research challenges from a
Software defined networking (SDN), among other emerging network management perspective. Throughout the paper the
technologies, holds the promise of 5G telecommunications terms applications, services, customers, sinks are used inter-
networks [3], [4]. It separates traffic management (the con- changeably for ease of presentation.
trol plane) from traffic forwarding (the data plane), enabling
centralized control and maximized agility [5]. In the control II. MTC SERVICE IN A NUTSHELL
plane, SDN manages network resources and controls network
traffic globally through one or a few SDN controllers. Based A. Traffic characteristics
on the status information from individual network elements MTC traffic is characterized as low-rate, small-packet,
and the overall traffic requirements, the SDN controllers make delay-tolerant traffic [8], [9]. In many MTC applications, a ma-
traffic control decisions by solving a traffic engineering (TE) chine transmits at a low frequency, e.g. one transmission every
problem; according to the TE solution, they provide control a few minutes or even hours in smart metering applications,

978-1-5090-1701-0/16/$31.00 ©2016 IEEE


Service request

and a small amount of data per transmission, e.g. tens of bytes (including VCSF-FG) MTC service
customer

for meter reading. Frequent or delay-sensitive transmission


happens only in emergency situations, for instance, upon VOSF-FG templates
Selecting VOSF-FG

detection of voltage surge in the electric power grid. MTC


traffic often may arrive in batches because of high machine VOSF information

Generating VNF-FG
density and correlated transmission. Correlated transmission (augmenting VCSF-FG with VOSF-FG)
VNF-FG

is normally due to machines detecting the same or correlated NFVI information

events or responding to the same data query, which also leads Creating IC-VNA
(deciding FE-Comp, FE-PoP and FE-LL)
IC-VNA

to traffic content correlation. Traffic information

SDT
In low-rate, small-packet MTC communications, a session
no longer presents itself as a thick flow of packets, albeit Fig. 1. A logical view of SDT functionality
possibly with a long time span. Instead, it is a series of single
or few packets per sessions, i.e. thin flows, scattered far apart
in time. Under this condition, TE’s per session traffic splitting factors, due to cascading failures attributed to the lack of
decision is meaningless since splitting does not happen at the efficient real-time diagnosis of loading in the electric power
packet level, and simple path routing may very likely perform grid [12]; and it led to a $10 billion social cost. To avoid
as well as multi-path TE. While individual machines have network congestion and reduce response time, in-network
low transmission rate (leading to a very small, less than 1%, processing should be enabled, as close to the radio edge as
volume share [10]), the total MTC packet count present in possible. On the other hand, it should not be too close to the
the network can be huge. It is anticipated that MTC traffic radio edge such that it can be fed with a sufficiently diverse and
accounts for approximately 10% of the total volume in terms wide range of information for making effective application-
of packets per second. Because every packet in transportation layer decisions.
occupies some network resources, e.g. CPU cycle at routers,
for packet header processing, it is more costly to convey data III. I NFORMATION -C ENTRIC V IRTUAL N ETWORK
via small-sized packets than through large-sized ones. In response to massive machine deployment, traffic corre-
lation and in-network processing desirability, we propose a
service-customized, information-centric virtual network archi-
B. Need for in-network processing tecture (IC-VNA). The architecture provides a TE framework
MTC services may involve downlink communication to that allows the network operator to define an on-demand and
configure machine operation for the purpose of business logic service-specific data plane to enable more efficient use of
enforcement, machine maintenance, and/or event response. network resources and to satisfy QoS the QoS requirements
When responding to an event, downlink MTC traffic may of the MTC service customers.
possibly have a strict delay requirement, depending on the An IC-VNA is a logical data plane topology customized for
nature of the event. To reduce response delay in traditional an MTC service. It includes service traffic sources, service
MTC networks, in-network processing and local reaction capa- traffic sinks and functional elements (FEs) as nodes and
bility have been introduced [11]. While the machines are inter- directed logical links between the nodes. A service traffic
connected in an ad-hoc manner, the network is connected to source can be an MTC device, an MTC device proxy, or a
the outside world through a few gateway (or sink) nodes. Raw base station serving MTC devices (proxies). A FE is a logical
machine data on its way to the sink nodes is aggregated and node used to host virtual functions. A virtual network function
processed at some internal aggregation points, which are either (VNF) is either a virtual customer-specific function (VCSF)
fixed or dynamically selected according to the temporal and or a virtual operator-specific function (VOSF) and can be
spatial correlation of nearby machines traffic. Data forwarding duplicated at multiple FEs. A VCSF is service-dependent and
paths are planned through the aggregation nodes points with reflects service business logic, while a VGNF is offered by the
respect to traffic correlation for improving data processing network operator to all or some MTC services for empowering
efficiency and effectiveness. The data processing functionality the networking process. One such VOSF that is necessary for
is fixed for each data aggregation point. all MTC services is serving gateway function (SGW). The FEs
When carrying MTC traffic, current telecommunications hosting a virtual SGW function for a specific service are called
networks treat machine reports as independent messages and virtual service-specific serving gateways (v-s-SGWs) [4]; they
deliver them to MTC application servers [8]. The application are connected with the service traffic sources through logical
servers extract and analyze the information carried by the links. Another optional VOSF is traffic trunking function which
messages and, when necessary, respond to machines and/or to can be used to mitigate the small-packet-count and large-
detected events. In other words, application-layer in-network packet-count problems. The primary mechanism used to create
processing is not offered. As the application servers sit on the an IC-VNA is software defined topology (SDT) [4] whose
other side of the network, the reporting-response cycle can specifics are given in the following.
experience large or unacceptable delay especially in loaded
network scenarios, which may cause huge economical and/or A. Software defined topology (SDT)
social loss in critical applications. For instance, the August In the ETSI NFV management and orchestration (MANO)
2003 electrical blackout in North America was, among other framework [13], VCSF and VOSF are not differentiated, and it
Service request Service request
MTC service Other input MTC service Other input
is assumed that for each virtual network service a pre-defined customer customer

VNF forwarding graph (VNF-FG) is required. The VNF-FG,


SDT-1 SDT
which describes a topology of all necessary VNFs, is supplied
to the NFV-MANO for determining the points of presence IC-VNA VNF-FG (update) FE-LL
VNF-FG
FE-Comp & FE-PoP

(PoPs) that hosts each of these VNFs. As illustrated in Fig. 1, SDT-2 (as part of NFVO)

NFV-MANO
FE-LL NFV-MANO

the SDT technology may leverage this framework to augment FE-LL FE-LL

a partial forwarding graph of VCSFs, denoted as VCSF-FG,


SDN (TE) SDN (TE)
with appropriate VOSF-FG templates to generate a complete
VNF-FG. VOSF-FG templates are pre-defined and maintained Provision FE-LL Instantiate FE Provision FE-LL Instantiate FE

by the network operator and can be updated on demand, as Data plane Data plane

the network evolves.


SDT further converts the VNF-FG to an IC-VNA, which (a) (b)
comprises defining function composition of FE nodes, place-
ment of FE nodes in the network, and logical links (including Fig. 2. IC-VNA creation and instantiation. (a) SDT partially residing in
NFV-MANO; (b) SDT residing outside NFV-MANO.
respective QoS requirements). We refer to these SDT decisions
as FE composition (FE-Comp) decision, FE point of presence
(FE-PoP) decision, and FE logical link (FE-LL) decision, VNF-FG generation may be intertwined with IC-VNA cre-
respectively. An FE can only be placed at an NFV-capable ation. If VNF-FG generation and IC-VNA creation are two
node, known as network function virtualization infrastructure separate processes, the later needs to feed back its result to
point of presence (NFVI-PoP) and referred to as NFVI node the former for potential VNF-FG adjustment; otherwise, they
in this paper. In the presence of multiple MTC services, SDT need to be carried out jointly, for example, through a single
jointly creates IC-VNAs for all services by taking into account optimization process.
of the service requirements, NFVI information and traffic 2) IC-VNA creation and instantiation: Once a VNF-FG is
information. When cross-service function sharing is allowed, generated, it can be input to the NFV-MANO. In this case SDT
node overlapping may appear between the IC-VNAs. is partially residing inside the NFV-MANO, for example, as
1) VNF-FG generation: A forwarding graph (FG) defines part of the NFV-Orchestrator component, to fulfill functionality
graphical constraints among virtual functions, namely, how of creating IC-VNA. Then, the NFV-MANO instantiates FEs at
they are inter-connected. Note that a single VOSF added by the selected NFVI nodes and interacts with the SDN (TE) for pro-
network operator can be viewed as a single-function VOSF- visioning FE-LLs. Alternatively, SDT may reside completely
FG. Then, VNF-FG generation generally comprises selecting outside the NFV-MANO. In this case, VNF-FG is an internal
VOSF-FG templates according to service request and augment- result of SDT; SDT outputs its result to a simplified NFV-
ing the VCSF-FG with the selected VOSF-FG templates such MANO, which may be the full IC-VNA or a partial IC-VNA.
that the overall graphical constraints are all satisfied. In some The NFV-MANO still takes the responsibility to instantiate
cases, template selection is done by classifying services and FEs. But, it may or may not need to interact with the SDN
pre-defining a template for each service class. For example, (TE), depending on if it receives FE-LL information from the
a VOSF-FG template containing virtual mobility management SDT. If the SDT does not provide FE-LL information to NFV,
entity (MME) function can only be used for mobile services. it will interact with SDN for FE-LL provisioning. Figure 2
In some other cases, the VOSF-FG has to be decided on the fly shows the two cases of IC-VNA creation and instantiation.
with respect to some network dynamics. Further elaborations
are provided in subsequent sections after introducing concepts
related to load-oriented VGNF graphical policies. B. Engaging customer
The graphical constraints defined by a FG can be broken IC-VNA decision is made on a per-service basis. The
into primitive VOSF graphical policies. SDT relies on these decision making is triggered by service requests submitted
policies during VCSF-FG augmentation. A VOSF graphical by MTC service customers. A service request includes two
policy regulates how a VOSF can be connected with other parts namely, service description and service requirement. The
virtual functions in the resultant VNF-FG. A given VNF first part depicts service traffic and service function. Service
may be subject to multiple such policies. Primitive VNF traffic description includes, for example, how the machines and
graphical policies may include linkage policies and ordering traffic sinks are distributed and what their traffic looks like
policies. A linkage policy defines whether a first VNF should (mean rate, maximum rate, transmission interval or schedule,
be linked with a second VNF, directly or indirectly, and in etc.); service function description covers, for example, impact
what direction. It may force both VNFs to be selected. For on traffic rate, VNF-FG, function instantiation constraints, etc.
example, a linkage policy for a virtual MME function and a The second part of a service request describes desired service
virtual SGW function says they must be directly linked in both traffic quality (e.g. minimum data rate, maximum delay, etc.),
directions. An ordering policy specifies the appearance order desired service function quality (effectiveness and efficiency)
of a VOSF with respect to the other VNFs along a function and how service function quality should be measured. Service
chain in a VNF-FG. For example, a virtual SGW function must function quality measure is function-dependent. For instance,
appear between a virtual eNodeB function and a virtual PGW an event detection function quality can be measured by false
function. Other polices are possible and are not discussed here. alarm rate, while a local reaction function can be evaluated by
MTC service Service request (update) SDT performance Service quality NFVI & traffic
customer Service information
measurement measurement information

SDT
IC-VNA FE node information,

NFV-MANO
estimated service quality, etc.

VCS
Fcontro
IC-VNAs Per-service IC-VNA link
SDN (TE) provisioning config
l and SDT Coordinator SDN (TE)
SDT configuration
Control plane
config

Per-service SDT checkpoints


- location, configuration
uratio

Per-service traffic control mechanisms


Control instructions
Devic

- location, configuration
n
e tra

NFV-MANO
nsmis

VNF1 VNF3
sion

VNF2
Control plane
sche

Control instructions
NFVI node
(including commands of instantiating SDT checkpoints and traffic control mechanisms)
du

Data plane
le (u
pdat

VNF 1 Traffic shaping


Wireless connections
e)

VNF 2 Traffic prioritization


Other traffic control
VNF 3
SDT mechanisams
Device domain checkpoint
NFVI node 1 NFVI node 2
Data plane

Fig. 3. Engaging customer Fig. 4. Adaptive IC-VNA realization

response delay.The procedure of IC-VNA creation is illustrated needs and to configure the application functions on the IC-
in Fig. 3. VNA FE nodes to reflect its latest business logic. When a
As shown in the figure, the network operator handles an service requirement changes, the customer issues a service
MTC service request through an SDT controller component, request update to the network operator, which then adjusts the
which customizes an IC-VNA for the service based on the current IC-VNA or re-creates a new one for the service. It is
service information and the information on network status up to the customer to decide whether and how to apply the
and NFVI node resource status. The SDT controller may also recommended transmission schedule. The customer may up-
decide for the service a recommended machine transmission date its service requirement based on application performance.
schedule that can result in achieving best network performance. For example, when the customer finds that in certain areas
Afterward, the network operator (SDT controller) updates the some urgent events cannot be effectively handled through in-
service customer with the IC-VNA node information (loca- network processing, it specifies the related areas as critical in
tion or network address, and respective functions), estimated a service request update. The SDT controller may then add a
service quality, and the recommended transmission schedule v-s-SGW, associating all the machines in those areas to the v-
(if one exists). Meanwhile, SDT controller also informs the s-SGW and instantiate an event detection function within the
NFV component about the IC-VNA FE node information v-s-SGW. Because the new v-s-SGW gathers a wider range
and the SDN component about the IC-VNA link information. of information related to the corresponding areas, it is able to
Following this, the NVF component instantiates the logical detect local events promptly. This closed-loop feedback allows
nodes while the SDN component provisions each logical link an adaptive IC-VNA from application perspective by engaging
with physical resources through TE, as elaborated below. the customer, enabling an elastic virtual edge to MTC services
Each FE node sends processed data to the next FE node that scales in and/or out according to the services’ need.
in the IC-VNA along a directed logical link. Traffic volume In conjunction with the customer-driven IC-VNA adaptation
and correspondingly, the rate may change after traffic passing described above, network-driven adaptation may be carried
through an FE node, depending on the virtual functions hosted out to enhance network system performance. As illustrated
at the node. For statistical functions such as computing the in Fig. 4, it involves an SDT checkpoint component and a
average or maximum temperature in environmental monitoring coordinator component. The SDT checkpoint is distributively
applications, traffic rate drops, and for functions like encryp- deployed at some, or all IC-VNA nodes to monitor SDT
tion, traffic rate often goes up. When multiple virtual functions performance, i.e. IC-VNA link quality (delay, delay jitter,
are called upon the traffic, they may balance out their effect throughput, etc) and stores the performance measurement in a
on the traffic rate. With SDT being engaged, SDN-TE does database. The coordinator reads per-service SDT performance,
not provision end-to-end service traffic flows but on IC-VNA service (traffic) quality (available from an end-to-end traffic
links, each being viewed as an individual flow. An IC-VNA monitoring component), service information, and network and
link is provisioned according to the link’s QoS requirements, NFVI node status (provided by a network monitoring compo-
the outgoing traffic characteristics measured or estimated at nent). Based on the information, it adapts the service-specific
the source end of the link, and the provisioning configuration SDT configuration (e.g. SDT optimization input parameters)
(e.g. number of physical paths per logical link). and IC-VNA provisioning configuration (e.g. link weight, link
capacity, number of implementation paths per link, etc).
The coordinator informs the new configuration information
C. Adapting to dynamics
to the SDT controller and the SDN component. Then, the SDT
The MTC service customer receives from the network controller updates or re-creates the IC-VNA for respective
operator IC-VNA FE node information, estimated service services accordingly, and the SDN component applies the
quality,and possibly a machine transmission schedule recom- adjusted configurations in subsequent TE optimization. The
mendation. This enables the customer to adjust its service coordinator component also determines whether to instantiate
an SDT checkpoints at some additional locations (if possible, connections among the selected subset of nodes such as to
to gain more performance insight for better coordination), or guarantee service traffic quality (statistically) through subse-
remove some existing ones (to reduce monitoring overhead quent IC-VNA provisioning with best resource utilization. The
when SDT performs well). It enforces the decision through the separation of architecture creation and provisioning also con-
NFV component. When necessary, through the NVF it installs tributes to the problem difficulty, while service traffic quality
or removes traffic control mechanisms such as traffic shaping guarantee is required. It involves additional research effort
and traffic prioritization on selected IC-VNA nodes for each to formulate the problem (objective function and constraints)
MTC service, to offer fine-tuned service performance. such that performance guarantee is achievable. The underlying
problem becomes even more challenging when considering
multiple services and cross-service optimization.
IV. D EPLOYMENT AND CHALLENGES
In addition to the IC-VNA creation problem, the coordina-
In real-world deployment, a geographic division (e.g. squar- tion problem for IC-VNA provisioning is a clear challenge.
ization) agreement is generally made between the network The challenge exists in that TE provisions individual IC-VNA
operator and MTC service customers. An MTC customer links while lacking a complete picture of end-to-end service
specifies in its service request the machine deployment at- traffic quality. When the network is loaded, coordinating IC-
tributes which may include specifications such as squares and VNA link provisioning within the same service and across all
per square machine count, machine traffic distribution, and services is crucial to providing service traffic quality guarantee.
per square data rate. As machines automatically associate to Traffic prioritization may take place at different levels, i.e.
base stations according to their location, the network operator service level, IC-VNA link level, and physical link level. It
calculates for each base station service traffic characteristics is an important traffic control technique to tune service traffic
and considers the base station as a virtual machine. It then quality. As part of IC-VNA provisioning, it is also a challenge
creates an IC-VNA for the service according to the virtual to coordinate IC-VNA link level traffic prioritization with the
machines. The virtual machines’ traffic characteristics can be traffic prioritization at other levels. Combining traffic shaping
updated over time through service traffic measurement done at at different places with traffic prioritization may offer improved
respective base stations. When necessary, the SDT decision can overall performance, however this comes at the expense higher
be updated or re-generated using the new information. When complexity in the coordination problem.
an IC-VNA logical node wants to communicate with machines
as a result of in-network processing, it generates traffic through R EFERENCES
the service function and the traffic is transported in downlink [1] R. Lu, X. Li, X. Liang, X. Lin, and X. Shen, “GRS: The Green, Reliability, and Se-
curity of Emerging Machine to Machine Communications,” IEEE Communications
to the machines. Magazine, vol. 49, no. 4, pp. 28-35, 2011.
As an IC-VNA FE node will have large processing load [2] 2.1 billion M2M devices in 2021, http://www.analysysmason.com/About-Us/News/
(gathering and/or processing machines traffic), the hosting Press-releases1/M2M-forecast-2012-PR-Jun2012/
NFVI node needs to have sufficient resources (bandwidth, [3] 5G: A Technology Vision, Huawei Technologies, April 2013. http://www.huawei.
storage, CPU power). Normally, in a high network utilization com/ilink/en/download/HW 314849
[4] H. Zhang, S. Vrzic, G. Senarath, N.-D. Dao, H. Farmanbar, J. Rao, C. Peng, and
scenario only a subset of the NFVI nodes are capable of H. Zhuang, “5G Wireless Network: MyNET and SONAC”, IEEE Network, vol.
hosting the IC-VNA nodes of a given MTC service. While this 29, no. 4, pp. 14-23, 2015.
may reduce the solution search space in identifying the host [5] S. Sezer, S. Scott-Hayward, P.-K. Chouhan, B. Fraser, D. Lake, J. Finnegan,
candidates for the particular MTC service and hence, enabling N. Viljoen, M. Miller, and N. Rao, “Are We Ready for SDN? Implementation
Challenges for Software-Defined Networks,” IEEE Communications Magazine, vol.
swift establishment of the IC-VNA, it may not result in the 51, no. 7, pp. 36-43, 2013.
global optimality of network resource utilization. This scenario [6] W.-C. Liao, M. Hong, H. Farmanbar, X. Li, Z.-Q. Luo, and H. Zhang, “Min Flow
can be more pronounced especially in the case when the NFVI Rate Maximization for Software Defined Radio Access Networks,” IEEE J. Selected
Areas in Communications, vol. 32, no. 6, pp. 1282-1294, 2014.
nodes are concurrently serving multiple MTC services, where
[7] H. Hawilo, A. Shami, M. Mirahmadi, and R. Asal, “NFV: State of the Art,
optimality may only be achievable through global reordering Challenges and Implementation in Next Generation Mobile Networks (vEPC),”
of all the services and their corresponding IC-VNA FE nodes. IEEE Network, vol. 28, no. 6, pp. 18-26, 2014.
For a particular MTC service such as vital signs monitor- [8] T. Taleb and A. Kunz, “Machine Type Communications in 3GPP Networks:
ing in an e-health application which requires ultra reliable Potential, Challenges, and Solutions”, IEEE Communications Magazine, vol. 50,
no. 3, pp. 178-184, 2012.
communications, it may be necessary to introduce a certain [9] M. Zubair Shafiq, L. Ji, A.X. Liu, J. Pang, and J. Wang, “A First Look at Cellular
level of redundancy in the network by duplicating the same Machine-to-Machine Traffic Large Scale Measurement and Characterization,” Proc.
application functions in different NFVI nodes. In this regard, SIGMETRICS, 2012.
determining sufficient level of redundancy for satisfying the [10] M2M application characteristics and their implications for spectrum, Machina
Research, 2014
service reliability while minimally impacting the performance
[11] I. Stojmenovic, “Machine-to-Machine Communications With In-Network Data
of other services is of paramount importance. Aggregation, Processing, and Actuation for Large-Scale Cyber-Physical Systems”,
Our solutions require the research and development of SDT IEEE Internet of Things Journal, vol. 1, no. 2, pp. 122-128, 2014.
technology for creating and provisioning service-customized [12] Final Report on the Implementation of Task Force Recommendations,
U.S.-Canada Power System Outage Task Force. http://energy.gov/oe/downloads/
IC-VNAs. Given a set of NFVI nodes, a daunting challenge is us-canada-power-system-outage-task-force-final-report-implementation-task-force.
to select a minimum subset of nodes and optimally associate [13] Network Function Virtualization: Architectural Framework, ETSI, 2013.
the machines to them. This is followed by jointly defining http://www.etsi.org/deliver/etsi gs/NFV/001 099/002/01.01.01 60/gs
the data processing functionalities and creating the logical NFV002v010101p.pdf.

You might also like