China Communications • October 2016
31
and a 25.1% improvement in the UAD-related
consumed-resource revenue compared with
the benchmark, moreover it has a 94.6% con-
verge ratio to the optimal solution of the aver-
age latency per path.
The remaining of this paper is structured
as follows: Section II introduces the heteroge-
neous 5G network infrastructure. Section III
provides the mathematical model, and section
IV includes the embedding-layered-auxiliary
graph and the heuristic algorithm. Section V
shows the numerical simulation results before
concluding this paper in section VI.
HETEROGENEOUS * NETWORK
NFRASTRUCTURE
The requirement to interconnect end users
with remote DCs supporting features such
as ubiquitous access and end user mobility
introduces the additional need for wireless
networks seamlessly integrated with the opti-
cal DC infrastructure [10]. And to be more in
line with the actual business, the design of 5G
infrastructure must take into account both het-
erogeneity and geographical distribution.
On the other hand, the complexity of
IT applications arises from the distributed
manner, in which they are processed in met-
ropolitan and wireless domains, such as the
concept of mobile cloud computing (MCC),
where computing power and data storage are
moved away from mobile devices to remote
computing resources [11]. Two reasons why
applications are distributed are: 1) the lack of
common resources at one location to meet the
gigantic computational needs of IT applica-
tions, 2) the geographic-distributed nature of
enterprises to facilitate globalized businesses.
The geographic diversity that results from
discrete enterprise entities implies a need for
an underlying metro network to be able to pro-
vision high-end applications. As opposed to
traditional networking paradigms that provide
only connectivity to end users, in order to fa-
cilitate emerging IT needs, the network has to
be smarter to adapt diverse-functional service
communications.
V-FiNE serves demands of delay diversity
and multi-level, and also the coverage re-
quirements are different. Simultaneously, the
lack of service differentiation mechanisms for
PRELOHDQG¿[HGFORXGWUDI¿FDFURVVWKHYDU-
ious network segments involved, the varying
degrees of latency at each technology domain
and the lack of global optimization tools in the
infrastructure management and service provi-
VLRQLQJPDNHWKHFXUUHQWVROXWLRQVLQHI¿FLHQW
Thereby, we propose a novel resource slicing
strategy for substrate network and we split the
HMD-5GN into multiple diverse-functional
layers (e.g., the smart terminal & mobile cloud
layer and the automotive devices & telemed-
icine layer), each of which integrates multi-
ple kinds of special-property nodes to adapt
different service requirement as well as user
access density (UAD). Besides that, we calcu-
late each demand a unique resource-attribute
vector (DC computing resource, BS wireless
channel capacity, fiber spectrum), and we
map this vector into our designed six-quadrant
service-type-judgment auxiliary graph to judge
which functional layer the demand should be
embedded on. And correspondingly, we give
each embedding layer a dedicated mapping
algorithm.
To the best of our knowledge, this paper
is the first work addressing the virtual 5G
network embedding problem in a complete
heterogeneous and multi-domain 5G network
infrastructure, based on the service-oriented
resource slicing strategy and the quadrant po-
sitioning method. Our contributions are sum-
marized as follows.
1) We proposed a heterogeneous and
multi-domain 5G network model for virtual
5G network embedding.
2) The problem was formulated using the
ILP model that obtains the optimal solution.
3) We proposed a service-oriented substrate
resource slicing strategy and constructed a
layered auxiliary graph.
4) We proposed a demand classification
method and a layered heuristic.
5) Our proposed heuristic can achieve a
47.4% improvement in average blocking rate