=Paper=
{{Paper
|id=Vol-1990/paper-13
|storemode=property
|title=Approaches for Optimization Using Virtual Network Functions in Infrastructure of Virtual
Data Center
|pdfUrl=https://ceur-ws.org/Vol-1990/paper-13.pdf
|volume=Vol-1990
|authors=Irina Bolodurina,Denis Parfenov
}}
==Approaches for Optimization Using Virtual Network Functions in Infrastructure of Virtual
Data Center==
https://ceur-ws.org/Vol-1990/paper-13.pdf
Approaches for Optimization Using Virtual
Network Functions in Infrastructure of Virtual
Data Center
Irina Bolodurina1 and Denis Parfenov2
1
Orenburg State University, Department of Applied Mathematics, Orenburg, Russia
2
Orenburg State University, Faculty of Distance Learning Technologies, Orenburg,
Russia
prmat@mail.osu.ru
Abstract. The study proposed a classification and identification model
of virtual network functions based on the statistical properties of the flow
and defined a systematic approach to the selection of the optimal set of
attributes of the traffic flow. The approach applied in our investigation
for placement of virtual network functions allows to optimizing traffic
flows in virtual data center. It includes algorithmic solutions based on
neural networks allowing to identify network functions.
Keywords: multi cloud platforms · network function virtualization ·
software-defined networks · virtual data center
1 Introduction
Today, considerable volumes of data circulate in modern telecommunication net-
works. The data centers are the nodes of aggregation and flow processing. The
modern paradigm of the network calculation environment demands the introduc-
tion of adaptive and flexible solutions. New solutions need effective control of
traffic in networks and, at the same time, do not demand cardinal changes of the
existing infrastructure of a data center. Traditional data centers use the concept
of resources virtualization for various infrastructure facilities (network, comput-
ing nodes, systems of storage and others) to achieve the goal. The application of a
complex approach to virtualization is reflected in the architecture of virtual data
centers. This architecture was used for the placement of multi cloud platforms.
Multi cloud platforms use hybrid methods of virtualization based on software-
defined components. It enables to increase the efficiency of computing resources
use and, thus, to reduce the economic cost of maintaining the infrastructure of
traditional data centers. However, the conception of resources virtualization is
not quite effective. It allows abstracting the processed and transmitted data flows
from physical devices. [1]. But, nowadays, the problem of the effective placement
of key components of the virtual network environment in a multi cloud platform
is not solved.
One of the approaches applied in virtual data centers apart from the virtu-
alization of traditional objects of network infrastructure, is the use of software
Virtual Network Functions in Infrastructure of Virtual Data Center 113
realization instead of traditional hardware solutions, such as firewall, load bal-
ancer, NAT, routers and others [3]. In practice, such solutions are based on the
technology of network function virtualization (NFV). The NFV technology pro-
vides more flexible deployment and enables to control the virtual objects of a
multi cloud platform, which perform the roles of hardware network devices, more
effectively. As a rule, the NFV technology is applied together with the software-
defined network and enables to exercise adaptive traffic control. However, the
technology of network function virtualization has a number of disadvantages.
The main problem is the lack of effective methods of planning for placing virtual
objects in physical computing nodes. The review of research shows that existing
solutions for placing the NFV in the infrastructure of data center use the ap-
proaches based on virtual machines or containers [2]. The existing solutions do
not deal with resource intensity of each virtual network function and its func-
tional purpose for multi cloud infrastructure of a virtual data center. We have
developed the approach that allows us to cluster the existing virtual and physical
objects of infrastructure and, then, to place virtual network functions. The main
idea of our solution is to estimate the consumption of resources by each element
of the network. Besides, we will use the hybrid method of virtualization based
on the simultaneous use of virtual machines and containers to create a flexible
solution. It will enable to optimize the placement of the technology of network
function virtualization in the infrastructure of a virtual data center.
Our approach is relevant, since it represents the combination of two modern
innovative technologies in the field of the organization of network functioning
and virtualization of its components for resource and data flow control in the
software-defined networks based on the technology of network function virtu-
alization. The goal of our investigation is to improve the quality of service for
applications and services of the multi cloud platforms placed in a virtual data
center. Besides, we use the methods of intellectual data analysis to process in-
formation about the state and load of key objects as well as the flows between
network devices received from the systems of computing nodes monitoring in the
software-defined infrastructure of the multi cloud platform. It enables to receive
the consolidated assessment of the quality of service and to predict uninterrupted
operation and operability of the software-defined infrastructure of a multi cloud
platform and the entire virtual data center.
In the following sections we’ll describe the approaches developed by us di-
rected to a solution of the problem of optimization placement of virtual network
functions in the multi cloud environment of virtual data center. Section 2 gives
information on a condition of researches today and the existing approaches in
work with network functions. In Section 3 we’ll describe the methods and ap-
proaches applied in our decision, and we’ll also describe the main stages of its
realization. The neuronetwork model which is a basis for formation of cards of
placement of network functions in the multi cloud environment of virtual data
center is presented in section 4. In section 5 the algorithmic solutions and experi-
mental explorations which in practice describing an optimization task placement
of network functions. Conclusions is presented in section 6.
114 Irina Bolodurina and Denis Parfenov
2 Related Work
The variety of physical network devices of various vendors increases both capi-
tal expenditure and operational costs for the maintenance of data centers. The
technology of network function virtualization allows solving this problem by the
realization of network functions as software. The application of the NFV implies
the use of the technology of network objects virtualization, which function as
software and particular computing processes, or as complex infrastructures of
cloud computing instead of hardware solutions.
The group of scientists headed by Min Chen [3] analyzes the architecture
and mechanisms of the interaction between the technology of network function
virtualization and the software-defined networks. As noted in the research, if
the number of the users who launch applications in the multi cloud network
environment increases, there is a competition for resources. Besides, each user
request is described by the relevant requirements to network environment from
the viewpoint of productivity, safety, and the effective control of objects in virtual
infrastructure.
Scientists from Arizona State University [4] have studied a multi cloud sys-
tem. They offer an approach to the creation of network architecture based on
the NFV technology as alternatives for traditional hardware network devices.
However, this research does not solve the main problem concerning the methods
of NFV placement on computing nodes.
Apart from a problem of the effective placement of network functions, the
NFV technology has a number of disadvantages associated with the organization
of the coordinated control of the entire network infrastructure of a virtual data
center [5]. To solve this problem, scientists from University of Wisconsin-Madison
have proposed an approach, which is a framework of the OpenNF. This approach
provides the effective coordinated control of both the internal state of network
functions and the state of data transmission network. However, this decision
does not solve another important issue associated with the overall performance
and load on the controller and objects of network functions.
The scientists from Nation Chung Cheng University have investigated the
reduction of load on the controller to ensure the work of the NFV technology [6].
As a rule, in case of the software-defined network use, the controller classifies
the traffic received from the ports on network nodes to define a path to network
objects, which play a role of the virtual network functions. This process generates
large volumes of traffic in the plane of control. The authors offer the expansion
for the architecture based on the software-defined network to reduce the load
created by network traffic on the controller due to the use of the NFV technology.
The solution represents two-layer classification of traffic based on the OpenFlow
protocol. Network events are analyzed in the plane of data instead of the plane
of control.
The group of scientists headed by Hassan Jameel Asghar has developed a
scalable system to work with the technology of network function virtualization
in the multi cloud environment [7]. The authors have offered an abstract model
of the standard network function distributed between several cloud platforms.
Virtual Network Functions in Infrastructure of Virtual Data Center 115
The developed model is used as a basis for the SplitBox system. This system
enables to increase the speed of package processing considerably in comparison
with similar hardware solutions. An essential disadvantage of the concept is the
resource intensity of this system. The overhead costs of the deployment and
work of the Splitbox network functions demand the same quantity of computing
resources as in case of hardware network devices. This disadvantage neutralizes
the available advantages of the approach, because the problem of the effective
placement of network functions in the multi cloud infrastructure of the virtual
data center remains unsolved.
Thus, it is established that the technology of network function virtualization
and the existing cloud solutions on based on the software-defined infrastructure
of the virtual data center has a number of advantages, which enable to improve
the quality of service in data transmission networks. However, today, there are
adequate and effective solutions, that would enable to control the placement of
the NVF on physical and virtual computing nodes in the data center.
3 Methods and Approaches
Nowadays, neural networks are the most effective and high-speed method for
forecasting, parameter identification, clustering and classifications in various
fields of knowledge. Today, we see many successful examples of the application of
a neural network approach for the creation of intellectual information systems.
Besides, the advantage of the neural networks use is the possibility of adaptive
self-training with the use of additional methods of approaches. We have used an
iterative approach based on a group of methods associated with the optimization
of placement of virtual network functions on the objects of the software-defined
infrastructure of the virtual data center.
First of all we will present all network objects of the multi cloud platform
placed in virtual data center as a communication graph. The graph is based on
the topology the physical network switching. Each network object is the graph
vertex. It can be described by a basic set of parameters, which characterize each
element of the network and influence productivity. We have chosen the follow-
ing characteristics as parameters: volume of memory, volume of disk space, the
frequency and quantity of kernels, etc. Further, we will use these characteristics
as the input parameters acting as the training set for a neural network during
the study of data.
A multi cloud platform supports the placement of various applications and
services. Therefore, to identify the flows of traffic passing through infrastructure
facilities of the data center is an important task for the placement of virtual
network functions. In this research, we have used the method based on the
analysis of the known network ports for popular applications to obtain this
information. This method enables to make an integrated classification of traffic
flows; however, since there are non-standard network solutions applied in service-
oriented applications, there obtained data will not be enough for the effective
control of traffic flows. For a deeper analysis of traffic flows as a data source, we
116 Irina Bolodurina and Denis Parfenov
offer to use the method of decoding the protocols of communication based on the
analysis of contents of the transferred packages. However, since this approach
has rather high resource intensity, it will be used only at a low level of the
analysis, for more exact identification of traffic flows of similar applications.
The third method uses the sample approach based on the specific signatures
located in protocol heading for the identification of the application. The fourth
method is based on machine training. This method uses the accumulated data
obtained by the above-mentioned methods and applies to them the algorithms
of machine training to identify the applications based on characteristic packages
and the saved-up statistics of data flows. The advantage of this approach is that
algorithms can be trained in real time that will allow reconfiguring software-
defined infrastructure of virtual data center on the fly. Use of the proposed
solution based on an integrated approach to collection of data on the traffic
flows circulating in a multi cloud platform. It will allow optimizing placement of
network functions on computing nodes of virtual data center.
To achieve the goals of the research, we have created a neural network system
to predict the placement of network function virtualization in the multi cloud
environment of the virtual data center. This implies the consecutive implemen-
tation of a number of algorithmic and software solutions. First of all, the module
of data collection, which enables to receive the sets of primary data about the
state of the network infrastructure of the virtual data center, is implemented
for a neural network system. The obtained information is necessary for neural
network training and testing. The next stage is to use the obtained data to define
the optimum scheme for the placement of network function virtualization and to
carry out experimental approbation on a-priori known samples and the obtained
results. This will allow us to correct the sets of input data and to improve the
quality of obtained results at the neural network exit. The final stage is to test
the system using the examples, which are not included in the training sample.
This will enable to ensure the efficiency of the obtained results.
4 Model of Clustering and Identification Virtual Network
Functions
We have chosen Kokhonen’s network as a neural structure for modeling, since
it is the most efficient in the clustering and classification of objects. Another
important factor is the visualization of results; it enables to improve the under-
standing of the structure and character of data at early stages and to specify
a neural network model further. Due to the peculiarities of network function
virtualization, the support of classification in Kokhonen’s network can be used
to identify the uniform elements in network for further optimization of their
placement. Kokhonen’s network is trained by a method of consecutive approx-
imations. Starting from the initial placement of objects selected randomly, the
algorithm gradually improves it to supply the data clustering. Another advan-
tage of Kokhonen’s network is the opportunity to identify new clusters. The
trained network detects clusters in the training data and refers all the data to
Virtual Network Functions in Infrastructure of Virtual Data Center 117
certain clusters. If the network meets a set of data, which differ from any known
samples, it will independently reveal a new cluster of elements then. This feature
is very relevant, since it allows entering new network functions into the archi-
tecture of virtual data center without the actual change of algorithms of their
distribution on physical and virtual computing nodes.
The principle of creation of neural network system for optimization of place-
ment of network function virtualization in a multi cloud environment of virtual
data center is as follows. On the basis of the data obtained from systems of mon-
itoring of virtual data center we have select a number of criteria. That possible
to identify unambiguously the network function placed on computing nodes. Cri-
teria are formulated so that the answer could be presented in the binary form
that is 1 - “Yes” or 0 - “No”. On the basis of the obtained data the table which
moves on an entrance of neural network is formed. Also the vector of output
values has a similar appearance. Its components also have a binary appearance.
We developed an algorithm of training of Kokhonen’s network. Kokhonen’s
network consists of one layer of neurons. The number of entrances of each neuron
is equal to n - it is total of the possible characteristics peculiar to network
functions. The amount of neurons of m coincides with the required number
of classes into which it is necessary to break, corresponds to the number of the
unique network functions used in work of a multi cloud platform. The importance
of each of entrances to neuron is characterized by the numerical size called by
weight. Describe the training of Kokhonen’s network by steps.
Step 1. Initialization of network.
Small casual values are appropriated to weight coefficients of network wi,j , i =
1, n, j = 1, m.
Values a0 -initial rate of training and D0 - the maximum distance between
weight vectors (W matrix columns) are set.
Step 2. Presentation of a new entrance signal to network of X.
Step 3. Calculations of distance from an entrance X to all neurons networks:
i=1
X
N
2
dj = xi − wi,j , j = 1, m (1)
n
.
Step 4. The choice of neuron k of 1 ≤ k ≤ m, with the smallest distance of
dk .
Step 5. Control of scales of neuron of k and all neurons which are from it at
the distance which isn’t surpassing dN .
N +1 N N
wi,j = wi,j + aN xi − wi,j (2)
.
Step 6. Reduction of values aN , dN .
Step 7. Steps 2-6 repeat until then, weight won’t cease to change (or still
total change of all scales will become very little).
After training classification is executed by giving on an entrance of network a
vector to be examined, calculation the distance from it to each neuron with the
118 Irina Bolodurina and Denis Parfenov
subsequent choice of neuron with the smallest distance as indicator of the correct
classification. For training of a neuronet we have used the data obtained from
system of monitoring of virtual data center of the Orenburg state university.
For experimental research we have selected key network functions which using
in all data centers to build standard network solutions. This allowed us to test
the functioning of neural networks and to determine the correct identification of
virtual network functions in the real system. In table 1 examples of experimental
recognition of virtual network functions in a software-defined infrastructure on
virtual data center are given.
Table 1. Result of experimental recognition of virtual network functions
Virtual The number of The number of
network function instances right recognized
Router 20 19 (98%)
NAT 15 13 (94%)
Firewall 18 17 (93%)
On the basis of the obtained data it is possible to make a conclusion that
for the constructed neurosystem presented certain difficulties of recognition of
a number of network functions in view of their insignificant differences in the
chosen parameters. This defect can be eliminated by introduction of additional
criteria to initial model of neural network. Thus, application of the developed
neural network system gives the chance to correctly identify virtual network func-
tions in 93-98% that, promotes increase in efficiency the solution of a problem
of optimization for their placement in infrastructure of a multi cloud platform.
5 Algorithm of Optimization of Distribution of Virtual
Network Functions in Infrastructure of Virtual Data
Center
The presented model of identification of virtual network functions allows opti-
mizing their placement in virtual data center. We will perform optimization of
placement the network functions which found by using Kokhonen’s network in
virtual data center by the following criteria: the current load which created on
computing nodes; resource intensity of network function; quantity of the flows
passing through computing nodes. The main objective of placement of virtual
network functions is the choice of optimum quantity of the nodes which realiz-
ing required functionality as software solution. Thus, the problem of planning of
resources takes place. At the organization of dynamic topology in virtual data
center planning is particularly relevant. The created load of computing nodes
can change over a wide range for rather short intervals of time and depends on
the chosen type of placing specific network functions. The algorithms of the mon-
itoring of infrastructure of virtual data center, placement and start of network
Virtual Network Functions in Infrastructure of Virtual Data Center 119
functions are developed for the solution of an optimizing task. In comparison
with the available analogs the algorithm uses the heuristic analysis of streams
of traffic, and also their classification depending on type of network function.
The integrated algorithm has the following points.
Step 1. To execute identification of an arrangement of virtual network func-
tions concerning topology of network infrastructure of virtual data center.
Step 2. To estimate quantity of the started copies of each virtual network
function and to rank them in order of requirements of network infrastructure.
The requiring is estimated concerning traffic flows which transferred through
copies of virtual network function.
Step 3. To define a load of physical and virtual computing nodes for each
copy of function.
Step 4. On the basis of the data obtained on steps 1 and 2 to make comparison
of data and define virtual network functions which need to scaling, minimizing
or turn off. The basic criteria used as input data to define set of virtual network
functions are network records and events, data of the time of packages going
through a network object, time of packet input and output, memory loading,
the use of CPU, the intensity of dataflow, TTL.
Step 5. For virtual network functions which requiring minimizing or turn
off to provide reconfiguration of topology on the controller of software-defined
network and executed a stop and release of the occupied resources of virtual
data center.
Step 6. For virtual network functions requiring scaling and creating the max-
imum load on infrastructure to executed an assessment of a way of dislocation.
To distribute the functions which are most loaded network using a hybrid way
of placement (the containers developed in the virtual machine). Less loaded,
but requiring scaling network functions to transfer to an operating mode in the
virtual machine.
Step 7. Provide step migration of the virtual machine and containers with
containing network functions on the least loaded hardware computing nodes in
data center. As part of our research, we propose a solution based on a hybrid
approach and allowing seamless migration of containers within a single network
space. The developed solution is based on a combination of two approaches to
resource virtualization. All virtual network functions are deployed as form of a
universal virtual machine. This virtual machine containing a set of containers.
Each containers consist only for one specified network functions. When the net-
work structure changing, the data on the content of the containers is replicated
to the required compute node of the software-defined infrastructure of the virtual
data center. This approach allows us do not interrupt the network connections
in the process of changing the state of network devices.
The approach applied in the offered control algorithm of placement of virtual
network functions allows to consider a way of placement and to organize work
of virtual data taking into account the circulating flows of a traffic regulating at
the same time quantity of the started copies of each function.
120 Irina Bolodurina and Denis Parfenov
6 Conclusions
The investigation proposed a classification and identification model of virtual
network functions based on the statistical properties of the flow and defined a
systematic approach to the selection of the optimal set of attributes of the traffic
flow. The results show that the classification of the traffic flows in cloud enable
to improve the quality of service by 20-25% by reducing the response time by
using virtual network functions.
In the future work we plan to explore more numbers of types of virtual
network functions which used for traditional tasks, and for build non-standard
solutions. Also, in the continuation of the research we are planned to investi-
gate the work of the developed algorithmic solutions in a distributed network
environment, including several geographically remote data centers.
Acknowledgments. The research work was funded by Russian Foundation
for Basic Research, according to the research projects № 16-37-60086 mol a dk,
№ 16-07-01004 and № 17-47-560046, and the President of the Russian Federation
within the grant for state support of young Russian scientists (MK-1624.2017.9).
References
1. Bolodurina I. P., Parfenov D. I. A model of cloud application assignments in
software-defined storages // Journal of Physics: Conference Series — Vol. 803.
2. Bolodurina I. P., Parfenov D. I. Development and Research of Models of Organiza-
tion Distributed Cloud Computing Based on the Software-defined Infrastructure //
Procedia Computer Science — Vol. 103 — P. 569–576.
3. Yong Li, Min Chen Software-Defined Network Function Virtualization: A Survey. —
IEEE, 2015.
4. Ruozhou Yu, Guoliang Xue, Vishnu Teja Kilari, Xiang Zhang Network function
virtualization in the multi-tenant cloud. — IEEE, 2015.
5. Aaron Gember-Jacobson, Raajay Viswanathan, Chaithan Prakash, Robert Grandl,
Junaid Khalid, Sourav Das, Aditya Akella OpenNF: Enabling Innovation in Network
Function Control. — IEEE, 2014.
6. Ying-Dar Lin, Po-Ching Lin, Chih-Hung Yeh, Yao-Chun Wang, Yuan-Cheng Lai
An extended SDN architecture for network function virtualization with a case study
on intrusion prevention. — IEEE, 2015.
7. Hassan Jameel Asghar, Luca Melis, Cyril Soldani, Emiliano De Cristofaro, Mo-
hamed Ali Kaafar, Laurent Mathy SplitBox: Toward Efficient Private Network Func-
tion Virtualization. — IEEE, 2016.
8. Bolodurina I. P., Parfenov D. I. Methods and algorithms optimization of adaptive
traffic control in the virtual data center // 2017 International Siberian Conference
on Control and Communications - Proceedings. — IEEE, 2017.