“a survey of software defined networking” › research › survey of software defined... ·...

عشر ثامنالعدد ال

م 2020 – نيسان – 2تاريخ اإلصدار:

www.ajsp.net 5798 -2663: ISSN

1 Arab Journal for Scientific Publishing (AJSP) ISSN: 2663-5798

“A Survey of Software Defined Networking”

Researcher:

Shadi Hadla

Dr. Ghassan Saba and Dr. Mohammed Alchaita

Higher Institute for Applied Sciences and Technology, Damascus, Syria

Email : [email protected]

http://www.ajsp.net/

mailto:[email protected]





Abstract:

It became more difficult to handle the networks in a proactive and reactive way after the rising number

of smart devices and mobile phones. Software Defined Networking SDN is relied on separating data plane and control

plane. SDN is considered a rising technology which assures that it will find solutions for most of difficulties and challenges

in the way of developing networks at the present time .SDN has a widespread control plane(CP) for all tools of networking

of the network that facilitates configuring of devices on the fly.

The way of how Software Defined Networks developed to be one of the most wanted technologies of current times is

studied in this paper. Moreover, it discusses the architecture and working of all SDN planes, and it presents how SDN is

applicable in many areas and some of them have been shed light on. SDN faces several security threats in each of its

planes. At the end of the paper the major security challenges are given in detail.

INTRODUCTION Sizable development in Internet has been caused by the large scale use of smart phones .we live in a computerized society

where everything is linked and accessible from anywhere. More control over traffic to achieve a suitable cooperative

connection is led by the increased dynamic behavior of nodes. There is a need of such versatile devices require wireless

connections with high mobility.

Virtualization, agility, mobility, and wireless connection with changing network configurations make network process a

difficult task. It is a necessity for the networks to boost diversity and at the same time is able to manage massive flow of

data that come from everywhere. The direction of emerging trends in the world of internet and technology is towards

combination all devices regardless their size. The network ability of each object requires cooperation of various

technologies.

It is also necessary to ensure the privacy of users, devices and the data being sent. Conventional IP networks are fixed and

very hard to handle in spite of the limitless extension and adoption in every dimension. The conventional networks are

closely combined with constituents of hardware which make it difficult to configure them. At the present, wireless and

wired networks uphold diversity and heterogeneity just for devices that have identical network features or protocol which

leads to weak Quality of Service and Quality of Experience (Zeng et Al., 2015) It becomes hard for network operators to

incorporate most recent technologies due to fixed and rigid nature of conventional networks. Thus, it is not perfect for

emerging business ventures. A programmable automatic network architecture that is applied in a dynamic way can offer a

typical resolution. Against background of the above stated issues of conventional networks, Software Defined Networking

(SDN) has developed as a proper solution for most of the worries. SDN is a graceful network architecture that presents a

centralized monitoring technique across the domain of network (Adam, 2014). Disconnecting the control from the network

devices, enabling network operators to handle network in a dynamic way (Drescher, 2014). is the core of SDN, it has taken

from early researches during early 2000 (Zegura et Al., 2014). Moreover, SDN includes one centralized control plane for

entire underlying actual network base, and the approach of SDN relies on the disconnection of networking devises control

and data planes. The control plane is put on top of Network Operating System and is made through combining the control

mechanism of all networking devices The data plane is just applied to redirect the data packets across the network. The

single main function executed at hardware level is redirecting of the packets and most of control connected jobs like policy

implementation are carried out at software level. The abstraction made by SDN has resulted in its fame in the domain of

creative technologies, for it presented wanted flexibility and virtualization in various environments of network. Security

has not yet been actually taken into consideration, and it is still an open problem despite of promises of SDNs to network

infrastructure that will solve most of the pressing troubles that face conventional networks. Technology comes first and

safety follows it. Security concerns increased in the SDN with centralized network logic or control mechanism. It seems

that SDN will benefit almost all computer technology areas mostly mobile networks, wireless sensor networks, cloud

computing, large data analytics and others. The rest of this paper is organized as follows: section 2 will discuss the SDN

Evolution. Section 3 identifies the virtualization in SDN. Section 4 introduces SDN languages. Section 5 presents methods

used to make SDN scalable. Tools for SDN described in Section 6. Section 7 discuss the advantages of SDN. Section 8

describes SDN architecture. And section 10 present applications of SDN in various areas.

SDN EVOLUTION Programmable networks are a concept that is not modern rather it is similar to the early research. It has

appeared from thoughts that were applied in early telephone call networks. As network evolutes the traffic across networks

evolutes, too. In early conventional network it was a hard function to handle the network execution, dependability, etc.

Therefore, approaches that can offer better network management were vital important to find (e.g. how to handle the traffic

across each node in a network). In addition, conventional network includes combination of control plane and data plane,






and this makes network management hard. So it presents an idea to researchers of disconnecting control plane and data

plane. Consequently, we can develop networks that are resilient, scalable, and dependable. Open Flow is a new method

which assists SDN in changing the way network were formerly preformed. The former networking researches that lead to

scalable networking architecture called SDN is argued in this part.

2.1 Architecture

It is an architecture that is consisted of four planes: data, discovery, decision and dissemination (Myers et Al., 2005). It

presents network level goal to configure the forwarding plane {devices and suggests giving the decision plane a centralized

view of the network. When it comes to Dissemination, it offers the strong communication over every hardware devices in

data plane. Discovery offers capability to every hardware devices to discover its own resources and regional environment.

Data plane for redirecting packet across underlying hardware switches.

2.2 Active Networking

During 1990s, it appeared when researchers proposed the concepts of new programming in the conventional networks

which allow major opportunities of innovations (Zegura et Al., 2014). Early network architecture applied the programs

such as Global Environment for Network Innovation s (GENI) and Future Internet Design (NSF FIND).

2.3 Ethane Ethane enables the operator of network to use network broad policy. Ethane can be implemented in both hardware and

software (Freedman et Al., 2007). It also offers the communication between two end hosts through permitting the precise

permission .All policies examined in ethane rely on flows and header fields that define the flow depending on the type of

packet.

1. VIRTUALIZATION IN SDN Network virtualization gives us the capability to make logical connections and virtual networks that are

disconnected from the hardware devices. SDN is a normal stand to help virtualization, and virtualization is considered the

main characteristic for the prosperity of SDN. Today’s networks completely rely on network virtualization and this can be

possible just through the combination and integration of control and data plane.

Network virtualization offers every tenant in networks with its own network topology and control over packet traffic flow

(Keller et Al., 2013). In Table 1, various controllers from different developers that function on different platforms and most

of them are open source are presented.

There are different controllers that assist virtualization such as NOX, FlowN, Maestro (Cox et Al., 2010), Beacon, POX,

Floodlight and Onix. FlowN assists the container completely based

Virtualization and the standard API enables the controller to offer connection between the physical and virtual networks.

FlowN is designed as developed version of NOX and it makes SDN scalable through presenting the illusion of directing

the disconnected controller.

Table 1: Controllers support OpenFlow Serial No. Developer Controller Implementation/Platform Freely Available

1 Stanford Beacon java yes

2 BigSwitch FloodLight java yes

3 Nicira/Stanford FlowVisor C yes

4 Rice University Maestro java yes

5 Nicira NOX Python/C++ yes

6 Nicira POX Python yes

7 CPQD RouteFlow C++ yes

8 Independent Developer NodeFlow java yes

for every host. Directing the disconnected controller for each host is very expensive process. It is unachievable process.

Every tenant can direct its private policies on controller but hardware devices via SDN. The essential and centralized

network virtualization is applying the current resources such as many reasonable examples of network consisted of virtual

nodes (Rexford & Keller, 2010), (Casado et Al., 2010). FlowVisor (Casado et Al., 2010) is considered one of the essential






famous methods relied on SDN and resembles the virtual links and networks. Every tenant that directs controller includes

several domains in packet header such as source IP address, destination IP address and port number etc. FlowVisor not

completely used the virtualization. FlowVisor has the power to host several guest controllers i.e. for each slice there is one

controller, each of which disconnected from each other. In Table 2 the various controller functionality. VeRTIGO

(Salvadori et Al., 2012) is the extension of FlowVisor, and it includes completely all network virtualization features.

VeRTIGO communicates

Table 2: CSDN Controller Functionalitie Serial No. Controller Functionalities

1 OpenFlow

2 Support Virtualization

3 Network Functionality

4 Scalability

5 Programmability

6 Performance

7 Reliability

8 Security

9 Centralized monitoring& virtualization

with controller and fundamental hardware switches through control channel.

This control channel contains Node Virtualizer, Classifier, Internal Controller, Port Mapper, Storage and VT Planner.

Network virtualization is somehow similar to VMwares virtualization. Nicira

Presented in Feb, 2012 the first network virtualization platform (NVP). NVP gives us the platform that is abstracted from

the hardware devices. Moreover, it enables the extensible formation of the virtual network.

2. SDN LANGUAGES Conventional networks consist of a set of different hardware devices, and every one runs diffused algorithm and protocol

with rules and policies of access, topology information, routing policies and traffic controlling services etc. To offer

communication among the different network is extremely hard. We should change whole setting which is a process that

takes long time whenever we want to add or remove any device from network.

Furthermore, conventional networks are composed of the combination of control and data plane that constrains the

scalability and resilience. Nearly all controllers such as NOX (Gude et Al., 2008), POX, Maestro (Cox & Cai, 2010) , and

Beacon (Erickson, 2013) provide the programmable interface that allows programs to respond to network events such

arrival of packet, drop of packet, and link status update etc. Early network protocols and overlay networks apply reasonable

programming such as NDlog (Stoica et Al., 2005) , Overlog (Condie et Al., 2005). When new networking architecture

emerged, i.e. SDN, it offered the platform to improve the programmable interface where application works. SDN enables

programmers monitoring the behavior of the abstracted hardware devices via programmatic control. Flow-Log is a

programming language for controllers of SDN. It is also a declarative language. FlowLog (Dougherty et Al., 2013)

programs consist of physical table that is identical to relational database. Flow-Log is improvement of datalog. With Flow-

Log programs, it is possible to obtain the present state of the controller and also to get capacity to achieve progressive

change to the packet area. Flow-Log programs are datalog that is not repeated. FML (Flow Management language) is one

of the major logic programming for controller. FML is a data log based, and it focuses to a large extent on writing policies

and rules. It is a field with particular high level of language.

Moreover, It provides OpenFlow networks (Lakshman et Al., 2013) with rule based formalism. It also includes different

rules and policies that are in charge of making decision to permit or ignore the specific flow of packet. Pyretic (Walker et

Al., 2013) is considered new language by which

Programmers can define networks rules and policies at abstraction of high level, and the abstract packet model is the

central consistent of Pyretic. Another language is FRENETIC which is a high level language for programming different set






of network hardware devices and tools. FRENETIC is a collection of language with declarative query tone, functional

stream language, and particular language.

3. SCALABILITY AND SDN In SDN the world view of the network is presented though the centralized controller which has straightforward control on

the hardware devices in data plane (DP).It is recognized that the fundamental switches and the centralized controller are the

constituents of network. Moreover, to direct the packet that relies on the rules and policies accumulated in the flow tables is

the major task of the switch. The fundamental switches are managed by the centralized controller through setting up the

rules and policy. Packet processing rules and policies are installed in hardware devices in two ways:

- Reactive: is that switches choose their routes by themselves.

- Proactive: is when first packet comes through entry executor. Packet is sent by entry executor, or in other words, by

ingress port to the controller and then the controller determines which route switches have to pursue.

SDN becomes flexible and resilient when hardware devices are configured through programmatic control. Here a question

arises: how can we make SDN scalable? There are many choices of controllers such as NOX, FlowN, FlowVisor and

Beacon etc. All of them offer the programmable interface which assists a low-level and event based model in which

programs, depending on the choice that is made by the centralized controller, manage the arrival of packet by either

dropping the packet or redirecting it. Nox is the most prevalent controller among the controllers. We have to get more

effective option of controller like ElastiCon (Lakshman et Al., 2013) in order to make CP control plane scalable. The fixed

charting between the redirecting hardware devices and the smart controller is assisted by early controllers as the dynamic

charting between the fundamental switches and the controller is supported by ElastiCon. By preserving the packet traffic in

the data plane (Wang et Al., 2010) which is extracted from the controller in CP, SDN can be scalable. These switches are

sufficiently intelligent to take any decisions. The switches have to have the ability to make decision about the place to

where the packet has to be redirected in order to manage the traffic flow effectively. This can be though having permanent

software program installed in switches with silicon chips and inbuilt buffer storage. ASIC’s (Application Specific

Integrated Circuits) is the base of these switches. In addition, these switches also include CPU, which consists of the full

forwarding tables, in order to manage the data plane traffic (Guo et Al., 2012). The rules installed in the switches firmware

inform the switch where the packet has to be redirected when it comes though switch in a network. Firmware is inbuilt

software in switches that make communication with centralized controller possible. Moreover, the capacity to recover

quickly after failure is another factor which makes SDN scalable. FatTire is the programming language that assists quick

failure recovery. The capacity to endure and recover from failure is the most prevalent and basic demands of any network.

Identical demands are in need for SDN. In SDN, by centralized controller, any failure is recovered spontaneously. To

recover from failure in seconds the network has to assist various routing paths and have the ability to work with whatever

type of networking topology (Tourrilhes et Al., 2012) SDN can change the flow of traffic by programmatic control

whenever failure happens or switch has a problem.

4. TOOLS FOR SDN SDN is a new pattern which makes the development of networks easier. In this part we present an introduction to tools that

are based on SDN such as emulators.

6.1 Mininet

Mininet (Lantz et Al., 2010) is an emulator or imitator which enables network to be imitated on one computer system.

Mininet is container based emulator (CBE) and assists process level virtualization, which is smaller form of virtualization.

In addition, the emulated and imitated server manages logic codes but events. It configures and directs the identical things

we discover on real network such like links, switches, servers and packets. Emulator offers the made-up environment with

made-up traffic similar to the real and actual network environment. Netkit (Rimondini & Pizzonia , 2008), CORE (Kim et

Al., 2008) and trellis (Feamster et Al., 2008) are other examples of container based emulators CBE.






6.2 DieCast

DieCast (Vahdat et Al., 2011) is nearly the same of the behavior of real network and it applied very few protocols and

software configurations diffused across hundred of physical networks. DieCast does not give the concentration to scaling

of hardware resources like main memory and disk etc.

6.3 ModelNet

ModelNet concentrates on the emulation of sizable topology of networks and that is the major point which makes

ModelNet different from other emulators. In ModelNet we need few resources to emulate network.

It is a scalable imitator which offers a broad scale emulation of the network. The ModelNet contains five stages:

Create, Distill, Assign, Bind and run. The emulation in ModelNet relies on real time hence every packet comes across the

emulated network with identical time interval, with same delay and same rate of loss of packet as the actual network.

5. ADVANTAGES OF SDN Everything is developing but the networking architecture. Early networks were not made to face the quick changing

requirements and demands of the current times. SDN can cope with the problems of traditional networking architecture.

Since SDN give us the programmable interface, SDN networks can be effortlessly scalable as network grows. Networking

hardware devices are rising day after day which leads to network growth.

Therefore, to configure all hardware devices manually in traditional network is very monotonous and expensive function.

SDN is completely preformed in software. Thus, it supports programmable control to configure the hardware devices.

These are some benefits of SDN:

7.1 Flexible Traffic:

To programmatically configure hardware devices that results in SDN flexibility and more gracefulness is the primer

function of the smart controller. Also we are able to change the functionality of controller depending on change in traffic.

7.2 Global View:

SDN offers global view of network which makes it easy for us to facilitate the network configuration and controlling. It

also makes it attainable to the user who intends to use it.

7.3 Easily Programmable:

In SDN, since control plane is segregated with the data plane, control plane is easily programmable. We have the ability to

change the function of control plane through changing the logic programming in CP whenever we have to conduct

changes.

7.4 Programmatically Configurable:

SDN configures the hardware devices through programmatic control, for SDN is completely preformed in software. The

controller directs the hardware devices to install that packet redirecting policy by programmatic control when the controller

chooses to perform specific policy.

7.4.1 Platform for Innovation:

As SDN is a new networking pattern, the disconnection of data and control plane which gives the chances for innovation

and creativity of new services and applications.

7.4.2 Faster Failure Recovery:

SDN supports virtualization that offers end-to- end connection among links and node across network. Thus, failure

handling is much quicker. The controller automatically checks if there is failure and manages them precisely.






6. SDN ARCHITECTURE Computer networks are considered to be three separated functional planes which are the data plane, control plane and

management plane. The data plane function is to redirect data. The data plane represents the policies and protocols applied

by the network to fill up the routing tables. The management plane offers supplementary services such as networks

supplying and controlling. The data plane redirects data according to policy as control plane enforce. In the control plane

these policies are determined. The three planes function in a cooperative way. The conventional networks have gathered

control and data planes on every network device. Good work and performance are presented by such purely decentralized

approach with fixed control and data planes until now (Uhlig et Al., 2014) . There is a necessity of higher network

resiliency and trust which caused the development of SDN because of exponential development of networks. Software-

Defined Networking (SDN) is new pattern that is anticipated to address constrains actions of the present networking

practices and ensure optimum levels of performance.

SDN network architecture relies on following points:

1. The control plane is lifted up to a centralized controller. Segregation of control and data functionality.

Switches are straightforward redirecting elements. Separation of duties turns switches into forwarding hardware

with minimal burden of policing. Few costly controllers and many inexpensive switches lead to the network

infrastructure cost effective.

2. Integrated and reasonable centralized controller enforces the control logic on the network by. The centralized

control in a SDN clarifies configuration of networks and facilitates enforcement of rules and policies job.

3. It is potential to make an operating system for networks, much like the operating systems of computer systems,

through the controller which is the most important part of the SDN. Via

corresponding software component called Network Operating System (NOS), SDN controller distributes the

control logic acrossthe network.

4. NOS acts, like others operating systems, as a mediator which offers an environment beneficial to programmability

and abstraction.

5. Packets are redirected according to the rules equipped by the controller in comparison to destination based traffic

flow of conventional IP based networks. The redirection decisions rely on flow rules. When a packet arrives at the

entry executor of a switch, its header fields are corresponded with the flow ingresses in a table, and the

statistics/counters are updated and corresponding actions are taken if any ingress matches.

6. The programmable network (software defined) is the main characteristic which is accomplished through running

software application on top of the SDN controller.

7. SDN offers a world view of the whole network and carries together diverse kinds of fundamental network

behavior through permitting flow abstraction and a centralized control.

The speeded physical existence of the SDN controllers is not prevented by the centralized control in SDN. To get

programmable and centralized control logic is important because it is simpler and less error-prone network policies

with high level languages and software support. A control program is able to react automatically to unusual behavior

of the network. The world view of the network status supports flexibility and agility of the networks (Uhlig et Al.,

2014).

A. Application Layer

The application plane is the top layer of SDN and it gives a proper place to make applications and services.

The SDN applications are the software or programmable part of SDN which is applied to divide network behavior

and requirements with the SDN controller by northbound APIs.

B. Northbound SDN interface

The northbound interface abstracts the physical network infrastructure. Communication between controller and

application layer is abstracted through the northbound interface. There is no current criterion for northbound SDN

interface.






C. Control plane

Unified and logically centralized controller which lies in the control plane manages the network. The controller is the

heart of the SDN. Networks are much like the operating systems of computer systems, so the controller makes it

possible to create an operating system. The centralized control in a SDN facilitates configuration of networks and

makes enforcement of rules and policies an easy job. The controller shapes the flow table ingresses and sends those to

the SDN switches through Southbound Interface.

D. Southbound Interface

Communication between the controller and SDN switches are presented by the southbound interface. The Southbound

Interface provides the flow rules, generated by the controller, to SDN switches. OpenFlow (OF) is most famous and

widespread protocol applied to make southbound communication.

E. Data Plane

Data plane is the important network infrastructures. This layer includes of the redirecting devices that are in charge of

forwarding packets according to the flow rules offered by controller via Southbound Interface.

It obtains the information from switches and sends back to the controller. The data plane is accountable for enforcing

direction policies in the SDN hardware.

7. APPLICATION OF SDN IN VARIOUS AREAS When software defined networking (SDN) emerges, it has caused several chances and opened the door for several

inventions in diversity of networking fields. By means of network programmability and virtualization.

SDN promises to control the majority of the issues faced in present networking technologies. This part sheds light on some

of the areas which can take advantage of SDN.

A. SDN in Data Center

SDN and Network virtualization is the new trend in data centers due to the dynamic task of resources between tenants of

the data center and to the general public. The ever rising density of servers need for better networking speed and bandwidth

make it necessary for data centers to hold more and more information. A cost effective approach to overcome this issue is

to merge several data centers into a one data center with more ability, and virtualization with such larger physical density

would serve as one of the best solutions. Virtualization has lesser power requirements and makes efficient use of hardware

and is also able to create, remove and extend or decrease the applications or services in lesser time. There are organizations

that have begun implementing SDN into production although the SDN for data centers is in progression. Google has

implemented Open SDN for directing WAN connections, Microsoft Azure applies overlay creating tens of thousands of

virtual networks, eBay has implemented public cloud virtual networks with VMwares Nicira switches (Weidong et Al.,

2015).

B. SDN for WAN

There is no specific traffic redirecting decision whenever failover happens in network due to the lack of a global view of

the network. The global view allows network operators to see all paths and other abilities from one fixed location. In

comparison to conventional IP networks, SDN controller is able to offer such a central view of the network and compute

optimal paths (Weidong et Al., 2015). In same context, a framework of software defined cognitive wireless networking has

been introduced that uses SDN and Cognitive Radio Networking (CRN) to create new use cases for wireless networks.

C. SDN in Other Environments

In present times, IT industry addresses several issues associated to networking devices such as management, scalability,

protection and resiliency. In conventional networking all the complex functions are carried out manually which cause weak

management of networks. SDN can be

anticipated as a solution for most problems. Incorporating SDN not only clarifies the networks problems but it also

decreases the operational cost considerably. Some of the essential characteristics of SDN which make it interesting contain:

managing variety of devices, developed configuration, logically centralized control, global view of the network, granularity






and flexibility. With such qualities SDN is appropriate in most of the technologies for simplification and easier operation.

Some of the areas where SDN has demonstrated its use have been shed light on.

8. conclusion SDN give us the new method to accomplish networking. Fully programmatic functionality of SDN makes it flexible and

hence scalable. This paper depicts how SDN architecture is the appropriate approach to face the rapidly changing demands

of networking organizations and customers. In conventional networks it is impossible to attain scalability and resiliency.

The centralized controller is able to change the functionality of hardware devices by changing the routing policy through

programmatic control. There is no necessity to change the hardware set up with changes in demands. SDN is the latest

concept in networking industries. It enhances the scope for further development in networking. It is the future of

networking which permits us to build the cost effective and agile networks. As in SDN the whole reliance is in the

intelligent programmable controller that is useful for management and control purposes but it also has drawbacks, for

controller is the major aim for the attacks. Also virtual network in SDN does not have security layer to protect the network

where as traditional networks are secure because we have various choices to protect network such as firewall, IDS, worm

detector etc. SDN OpenFlow (Abdelaziz et Al., 2015) interface is also weak from security point of view. Also it has other

issues like whether the functionality will reside in the control plane or both (data or control plane).

References

[1] Zeng, L. & Xin, W. & Li, Y. & Jin, D. & Yang, M. & Vasilakos, A. (2015). Software-Defined and Virtualized

Future Mobile and Wireless Networks: A Survey, Mobile Network Applications, vol. 20, pp.418.

[2] Drescher, A. (2014) A Survey of Software-Defined Wireless Networks,

http://www.cse.wustl.edu/ jain/cse 574-14/ftp/sdwn/index.html.

[3] Zegura, E. & Rexford, J. & Feamster, N. (2014) The Road to SDN: An Intellectual History of Programmable

Networks. ACM SIGCOMM, Volume 44 Issue 2, 87-98

[4] Myers, A. & Hjalmtysson, G. & Maltz, D. A. & Greenberg, A. & Rexford, J. & Xie, G. & Yan, H. & Zhang, H.

and Zhan, J. (2005) A clean slate 4D Approach to network control and management. ACM SIGCOMM Computer

Communications Review, 35(5), 4154.

[5] Freedman, J. M. & Pettit, J. & Luo, J. & McKeown, N. & Shenker, S. & Casado, M. (2007)

Ethane: Taking Control of the Enterprise. SIGCOMM 07, 2731.

[6] Keller, E. & Rexford, J. & and Drutskoy, D. (2013) Scalable Network Virtualization in Software-Defined

Networks. IEEE Internet Computing,

[7] Cox, L. A. & Cai, Z. & Ng, T. S. E. (2010) Maestro :A System for Scalable OpenFlow Control, Rice University

Technical Report TR10-08,

[8] Rexford, J. & Keller, E. (2010) The Platform as a Service model for networking. In Proc. of USENIX

INM/WREN, San Jose, California,

[9] Casado, M. & Ramanathan, R. & Koponen, T. & Shenker, S. (2010) Virtualizing the network forwarding plane. In

Proc. of ACM, PRESTO, Philadelphia, USA,

[10] Casado, M. & Gibb, G. & Sherwood, R. & Parulkar, G. & Appenzeller, G. & McKeown and Yap, K. N. (2010)

Can the Production Network Be the Testbed?. In Operating system Design and Implementation,

[11] Salvadori, E. & Gerola, M. & Riggio, R. & Corin, D. R. & Pellegrini, D. F. (2012) VeRTIGO: Network

Virtualization and Beyond. In EWSDN, 24-29

[12] Gude, N. & Koponen, T. & Pfaff, B. & Pettit, J. & Casado, M. & McKeown, N. & Shenker, S. (2008) NOX:

Towards an Operating System for Networks. ACM SIGCOMM Computer Communication Review, Volume 38 Issue 3,

[13] Erickson, D. (2013) The Beacon OpenFlow controller.In Proc. HotSDN 2013.

[14] Stoica, I. & Loo, T. B. & Hellerstein, M. J. & Ramakrishnan, R. (2005) Declarative routing: Extensible routing

with Declarative Queries. In Proceedings of SIGCOMM,

[15] Condie, T. & Hellerstein, M. J. & Loo, T. B. & Maniatis, P. & Stoica, I & Roscoe, T. (2005) Implementing

Declarative Overlays. In Proceedings of SOSP,






[16] Dougherty, J. D. & Guha, A. & Nelson, T. & Fisler, K. & Krishnamurthi, S. (2013) A Balance of Power:

Expressive, Analyzable Controller Programming. HotSDN13,

[17] Lakshman, V. T. & Hao, F. & Dixit, A. & Mukherjee, S. & Kompella, R. (2013) Towards an Elastic Distributed

SDN Controller. HotSDN13,

[18] Walker, D. & Reich, J. & Foster, N. & Monsanto, C. & Rexford, J. & Cornell, P. (2013) Composing Software-

Defined Networks. 10th USENIX Symposium on Networked Systems Design and Implementation,

[19] Wang, J. & Rexford, J. & Yu, M. & Freedman, J. M. (2010) Scalable Flow-Based Networking with DIFANE.

SIGCOMM 10, New Delhi, India.

[20] Guo, C. & Miao, R. & Xiong, Y. & Lu, G. (2012) Using CPU as a Traffic Co-processing Unit in Commodity

Switches . HotSDN12,

[21] Tourrilhes, J. & Schlansker, M. & Kim, H. & Santos, J. R. & Turner, Y. & Feamster, N. (2012) CORONET:

Fault tolerance for Software Defined Networks. In Proceedings of ICNP. 1-2.

[22] Lantz, B. & Heller, B. & McKeown, N. (2010) A Network in a Laptop: Rapid Prototyping for SoftwareDefined

Networks. In HotNets. ACM,

[23] Rimondini, M. & Pizzonia, M. (2008) Netkit: easy emulation of complex networks on inexpensive hardware. In

International Conference on Testbeds and research infrastructures for the development of networks & communities,

TridentCom ’08, Pages 7:1f7:10, Brussels, Belgium, ICST.

[24] Kim, H. J. & Danilov, C. & Ahrenholz, J. & Henderson, R. T. & Danilov, C. (2008) CORE: A real-time network

emulator. In Military Communications Conference, MILCOM ’08, 1 IEEE, [25] Feamster, N. & Motiwala, M. &

Bhatia, S. & Muhlbauer, W. & Mundada, Y. & Valancius, V. & Bavier, A. & Peterson, L. & Rexford, J. (2008)

Trellis: a platform for building flexible, fast virtual networks on commodity hardware. In CoNEXT ’08, pages 72:1 f

72:6. ACM,

[26] Vahdat, A. & Vishwanath, V. K. & Gupta, D. (2011) DieCast: Testing Distributed Systems with an Accurate

Scale Model. ACM Transactions on Computer Systems 29, 4:1-4:48.

[27] Uhlig, S. & Rothenberg, E. C. & Ramos, M. V. F & Diego Kreutz, D. & Azodolmolky, S. & Verissimo, P.

(2014) Software-Defined Networking:A Comprehensive Survey.

[28] Weidong, W. & Hongyu, P. & Yinghai, Z. & Chaowei, W. & Gang, C.(2015) QoS-

guaranteed energy saving routing strategy using SDN central control for backbone networks, vol.5, pp. 92100.

[29] Abdelaziz, A. & Gani, A. & Khan, U. S. & Akhunzada, A. & Anuar, B. N. & Khan, K. M. & Hayat, A. (2015)

Secure and Dependable Software Defined Networks Journal of Network and Computer Applications- Elsevier

(Article in Press).


“a survey of software defined networking” › research › survey of software defined... ·...

Documents