WASEDA UNIVERSITY

Graduate School of Fundamental Science and Engineering

Master’s Thesis

Study of Named Node Networking

Implementation Designing

July 19th, 2016

Qi Xin

(5114FG16-0)

SATO Laboratory

(Professor Takuro Sato)

1

Acknowledgement

I would love to express my greatest gratitude to my supervisor, Prof. Takuro Sato, who

has provided me with encouragement and constant support during my study of master

course. I would like to give the thanks to Prof. Wataru Kameyama and Prof. Hidenori

Nakazato for the support on the study and research in ICN. I would like to thank Zheng

Wen and Li Zhu along with all the other respectful seniors in SATO Lab, who have

supported my research and study in Waseda University. I also would like to thank Jairo

Lopez for such great project and opportunity to perform research in the field of future

network.

Special thanks to Mr. Kazuhisa Yamada and Mr. Yuki Minami from Network

Innovation Laboratories, NTT, for the support in the VNode testbed.

2

Contents

Acknowledgement ........................................................................................................ 1

Contents ........................................................................................................................ 2

List of Figures ............................................................................................................... 4

List of Tables ................................................................................................................. 5

Chapter 1 ...................................................................................................................... 6

Research Background ................................................................................................................ 6

1.1. Future Network ........................................................................................................ 6

1.2. New Network Architecture ...................................................................................... 7

Chapter 2 ...................................................................................................................... 8

Introduction ............................................................................................................................... 8

2.1. Information-Centric Networking ............................................................................. 8

2.2. Introduce 3N .......................................................................................................... 14

2.3. Simulators .............................................................................................................. 18

2.4. VNode .................................................................................................................... 20

Chapter 3 .................................................................................................................... 22

Foundation of Evaluation ........................................................................................................ 22

3.1. Evaluation Method ................................................................................................. 22

3.2. File Descriptor NetDevice ..................................................................................... 23

Chapter 4 .................................................................................................................... 25

Experiment .............................................................................................................................. 25

4.1. Experiment Architecture ........................................................................................ 25

4.2. Experiment Environment ....................................................................................... 26

3

4.3. Experiment Results and Evaluation ....................................................................... 31

Chapter 5 .................................................................................................................... 33

Conclusion ............................................................................................................................... 33

Future Work ............................................................................................................................. 33

References ................................................................................................................... 34

Publication List .......................................................................................................... 35

4

List of Figures

Fig.1. Structure Compare of TCP/IP and ICN ........................................................ 9

Fig.2. Procedures of a ICN Request .......................................................................... 13

Fig.3. 3N Packet Structure .......................................................................................... 15

Fig.4. ndnSIM Components........................................................................................ 19

Fig.5. Overview of nnnSIM module .......................................................................... 20

Fig.6. VNode Servers’ Locations ............................................................................... 21

Fig.7. Simulated Link in ns-3 ..................................................................................... 24

Fig.8. Network Structure with FdNetDevice .......................................................... 24

Fig.9. Basic Tree Topology .......................................................................................... 26

Fig.10. Ideal 3N Experiment Configuration ............................................................. 28

Fig.11. Emulated Roaming Process ............................................................................ 29

Fig.12. Experiment Structure in VNode2 .................................................................. 30

Fig.13. Throughput of Lab Experiment..................................................................... 31

Fig.14. Throughput of Emulated Experiment with 1 switch ................................ 32

Fig.15. Throughput of VNode2 Experiment ............................................................. 32

5

List of Tables

TABLE.1. Interest Packet ......................................................................................... 11

TABLE.2. Content Packet ......................................................................................... 11

TABLE.3. PDU Types ................................................................................................ 16

TABLE.4. 3N Node Structure ................................................................................... 17

TABLE.5. Roaming Schedule in Lab Experiment .................................................. 27

TABLE.6. Roaming Schedule in VNode2 ................................................................ 30

6

Chapter 1

Research Background

After the invention and application of internet in the 1960s and 1970s. Widely use of

the internet impacts human society again and again by every revolutionary application

and evolvements in recent years. The internet has affected many aspects of our daily

like and society, improving and changing our lifestyle, work, communication and social

interaction. Everyone should not deny the fundamental role of the internet on our

society, our world.

1.1. Future Network

Nowadays, the internet is a complex combination of protocols that were developed in

the past. They were designed in the time when the technological development differed

very much from today. For the convenience of developing and using the internet in the

future, many improvements were submitted in the better foreseeing for the future

internet architecture. There comes the question, will the current TCP/IP stack still

fulfilling the needs of our information society in the future? Since this question was

proposed, there were many initiatives aimed to reshape the internet structure appeared

all around the world, these visions were called designs of Future Internet.

7

1.2. New Network Architecture

Considering the current situation on computing and communications via worldwide

network, is it possible to come up with a new design of the network architecture in

order to fulfill the short boards in today’s appliances. With the current network structure

mainly designed for TCP/IP, it is not aiming the information that states so important in

nowadays communication networks. Here comes to the proposal of Information-

Centric Networking (ICN) [1], the network architecture designed for content oriented

communication networking today. ICN network uses named contents instead of

addressed hosts applied in TCP/IP. Following the proposal of named content, many

projects, Named Data Networking (NDN), CCNx, took place worldwide in order to

implement and further discuss the form of ICN architecture.

8

Chapter 2

Introduction

In this section we are going to introduce several network concepts that have the vision

to improve the network structure in the future.

2.1. Information-Centric Networking

The Information-Centric Networking is a network architecture that aiming to create a

clean slate solution that seeks to address and improve some of the problems and

limitations in current network architecture. The fundamental concept of the basement

in ICN is to replace the host-based naming scheme in current network structure with an

information-based one instead. In order to understand how ICN works, it is important

to become familiar with the naming patterns, main components, forwarding mechanism,

along with the advantages ICN has over TCP/IP.

2.1.1. The concept

The current architecture of the Internet, which uses TCP/IP as its protocol of choice,

has been escalated from an academic network to a worldwide infrastructure that

connects large scale of information and great amounts of devices. Recent researches

9

have shown that there exists over 1 billion devices, 1 trillion of indexed internet pages,

and Exabyte of data being transferred worldwide every year [2]. Moreover, it is widely

seen the information distribution transfers great amount of data and clients interested

in obtaining contents whenever and wherever they might locate. The TCP/IP based

network architecture is not designed to fulfill the content distribution and is based in a

host-centric communication model. In another word, a user expecting certain

information could not be satisfied by only knowing what the content is, but also has to

instruct the server or host from which it can be provided.

Fig.1. Structure Compare of TCP/IP and ICN

ICN is designed to be the network architecture that delivers information by named

contents, so in this architecture, the content itself is the key of the scheme. By offering

names to the contents, ICN makes the contents address-free, no more limitations from

10

the eager of knowing locations. In another word, this network structure primarily

distributes information [3]. This makes difference from TCP/IP protocol, in which

contents need to be retrieved from specific addresses. The easiest way to understand

the meaning and differences between ICN and TCP/IP is to get in touch on what they

insist to allow in communications; for TCP/IP is the location and for ICN is the

content’s name. Meanwhile TCP/IP routes contents from the source to the destination

and ICN forwards them from the producer to the consumer.

The description in comparing IP and ICN could be generalized in to two hourglass

shaped charts, as shown in Figure 1.

2.1.2. ICN Packets

In the communication architecture of ICN network, there are two main kinds of packets,

content packets and interest packets. Table 1 and Table 2 shows the basic structures of

the packets.

The mechanism of the communication in ICN network is formed by these two kinds of

packets. In order to fulfill each request of desired data, the consumer has to send an

interest which contains the name prefix describing the desired content. As the interest

packet being sent out to ICN nodes, the nodes will respond with a content matching the

name prefix to the consumer, or on the other hand, forward the interest to the other

nodes to see if they own any valued contents.

11

TABLE.1. Interest Packet

TABLE.2. Content Packet

12

2.1.3. Data Structure

In ICN data structure there are three main elements, they are CS (Content Store), PIT

(Pending Interest Table) and FIB (Forwarding Information Base). These functionalities

are installed in every ICN node as default.

The CS provides a caching function for the contents in ICN nodes. There are two kinds

of contents that flow in the ICN network, self-generated contents and forwarded

contents. Either kinds of contents will be cached in the CS in order to respond to another

interest containing the same name prefix as the content generated or forwarded before.

The PIT provides the track back function to the ICN nodes. The action takes place when

an interest is being forwarded to other nodes. The interest’s receiving face would be

recorded for later tracking back and forwarding the content.

The FIB is acting to forward interests to any potential data producer. When an Interest

packet comes to the node, it will check the FIB for any pre-inserted forwarding

information to know which face should the interest be forwarded to.

13

Fig.2. Procedures of a ICN Request

Figure 2 shows the procedures that containing how the CS, PIT and FIB work with each

other. First the consumer node sends out the interest generated with a certain name

prefix. If the next node that does not hold the correct content receives the interest, it

will forward the content to next possible node after checking with the CS for content.

During the forwarding procedure, the node would have check the FIB to know which

14

face to forward the interest to. At the half way of the request, meaning the content

producer receives the interest, the content that matches the name prefix in the interest

will be transfer back all the way to the consumer depending on the PIT in every node.

2.1.4. Related Works with ICN

By the time Van Jcobson giving the popular speech of named content network in

Germany, many research groups started to perform their understanding on the ICN

network. There are many projects under development now like Named Data

Networking (NDN) [4] led by University of California. Simulators and real network

stacks are being tested and applied, like ndnSIM [5] and CCNx [6].

2.2. Introduce 3N

From what we have discussed above, the ICN network architecture proposed to cater

to the trend of network shifting from “host-centric” to “content-centric” networks.

However, in the appliances of mobile network, ICN architecture is not performing quite

efficient because during the roaming processes among access points, the moving

consumer node is still suffering packets loss in a great deal.

In order to fulfill the request of a seamless mobile network, Named Node Networking

(3N) [7] is introduced. 3N network concentrates on the content distribution, utilizing

15

the interconnected and communication packets for the purpose to reduce the node load

on particular content producers, meanwhile produces low packet delay instead of packet

retransmit on current generation of the network system.

2.2.1. 3N Architecture

To realize the goal of naming the nodes and utilizing the named nodes, a set of Protocol

Data Units (PDUs) is initialed. Table 3 provides the important information about 2

particular types of PDUs: data transmission PDUs and mechanism PDUs. The Data

Transmission PDUs hold the 3N names and works to inform any nodes in the

transmission path to forward the packets to their destinations. Mechanism PDUs holds

the instruction information that could manipulate ICN network layer to act in a

particular way to fulfill the functionalities of 3N architecture.

Fig.3. 3N Packet Structure

16

TABLE.3. PDU Types

PDU Types

Data Transmission PDUs

SO Includes Source node’s 3N name only

DO Includes Destination node’s 3N name only

Mechanism PDUs

EN Enrolls node and requires name

OEN Offers a name to an enrolling node

AEN Acknowledges the enrollment

REN Reenrolls and requires a new name

DEN Disenrolls from previous node

ADEN Acknowledges the disenrollment

INF Informs the obtaining new name

2.2.2. 3N Node Name Structure

Additionally, to the basic structures in the ICN network, 3N architecture has some new

features. The Node Name Signature Table (NNST) and the Node Name Pairs Table

(NNPT) are introduced. Table 4 shows the details.

17

TABLE.4. 3N Node Structure

3N native formation

NNST

Record the names, lease time of

3N name, and the 3N nodes

enrolled through connected nodes.

Maintains the mapping of 3N

names.

NNPT

Records new 3N names and old 3N

names, maintaining the mapping

between them.

2.2.3. 3N Roaming Mechanisms

There are three main mechanisms involved in the roaming procedures of 3N network,

enrollment, disenrollment and reenrollment.

In a basic network structure, e.g. tree topology, a node sends out EN packet to every

possible name producer to get the first name. When the EN packet hits a name producer,

the producer will reply a AEN packet to sign the node as a lower level node from itself.

Since then, the newly given name node is ready to perform in the 3N network.

For a mobile consumer that moves physically in the ICN network, in another word,

roaming from access point to another, the old name from previous node is no more

usable in the new area. During the roaming process, the disenrollment and reenrollment

18

mechanisms are used to help the mobile node to get new name and along retrieves

contents that requested to the last access point.

2.3. Simulators

Following are the introductions of some simulators involved in my master thesis.

2.3.1. ns-3

For the goal of realizing the evaluation of 3N network architecture, the first step is to

apply it to a certain network simulator. One popular choice is to implement on the

Network Simulator 3 (ns-3) [8]. Ns-3 is targeted mainly in the field of performing

research and educational. The developing goal of the ns-3 project is to provide network

researching with a preferred and open simulation environment.

The ns-3 fundamentally supports the architecture of both IP and raw packets

communicating networks. This is quite important for realizing 3N architecture along

with ICN.

2.3.2. ndnSIM

In the field of developing ICN network, the Named Data Networking Simulator

19

(ndnSIM) is well known as a project under UCLA.

The ndnSIM is a module in the ns-3 that implements NDN communication model as

shown in Figure 4. The implemented module uses separate classes based on C++ to

manipulate behavior of each network-layer entity in NDN. Also this modular structure

allows any modification or replacement in order to realize target network functionalities.

2.3.3. nnnSIM

The nnnSIM uses module to implement 3N functionalities into ns-3 simulator. The

design provides 3N functionalities along with pore NDN functionalities. The module

modified from ndnSIM v0.1 [9], together with basic NDN features. Figure 5 shows the

overview of the functionalities in the nnnSIM as in the module of ns-3.

Fig.4. ndnSIM Components

20

Fig.5. Overview of nnnSIM module

2.4. VNode

VNode [10], provided by NTT, manage to realize the deep programmability in the

visualized network, also that integrates computing resources. Services chained to

provide cooperative operations to the executable data processing function. The VNode

offers the chance to deeply program the virtualized network environment in order to

perform complex experiments. VNode distributes servers and network services, spreads

over the server in Japan. During the experiment on VNode, servers located in Fukuoka,

Osaka, Nagoya and Tokyo on JGN-X, shown in Figure 6, are used.

21

Fig.6. VNode Servers’ Locations

22

Chapter 3

Foundation of Evaluation

Evaluating a proposed network structure is a very important part in the research, with

the nnnSIM we can perform simulations without difficulty. However, on the other hand

of the next step, performing experiments in complex network environment is

considered to be vital part of the evaluation.

3.1. Evaluation Method

To realize the goal of evaluating 3N architecture in complex network environment, e.g.

physical network structures and virtualized network structures, 3N packets have to be

transmitted out of the box. In the field of ICN projects, there are already some

implementations like CCNx and NDNx, both of them have taken much time and

manpower to develop even just a prototype.

Considering the need of implementing 3N in a quick method, the better choice is to use

nnnSIM generated 3N packets on the network structures in a more direct way. The

tentative idea is to find a method to forward 3N packets onto the network stack and let

the host do the rest of the work. Thanks to the rich resource of the ns-3, a module which

could realize certain function was caught in sight.

23

3.2. File Descriptor NetDevice

In the modules of ns-3, the fundamental base of nnnSIM, there is one named File

Descriptor NetDevice (FdNetDevice) [11]. This module offers the chance to us

FdNetDevice class, which has the function to realize reading and writing package traffic

using a file descriptor that used by the user. The NetDevice file descriptor can be

managed to work with a TAP device, a raw interface, a user space process producing or

consuming data flow, etc. The system controller has the overall authority to define the

generation along with consuming to the outside.

Figure 7 shows the architecture of communication specs inside ns-3 simulator, this

refers to the limitation of the simulated link among the nodes. Meanwhile Figure 8

shows the structure of nodes communication between two hosts. The examples are

based on TCP/IP protocol.

With the help of FdNetDevice, we are able to forward nnnSIM generated packets to the

file descriptor and generates network traffic to host’s network devices to finally achieve

the goal of evaluate 3N performance under complex network environment.

24

Fig.7. Simulated Link in ns-3

Fig.8. Network Structure with FdNetDevice

25

Chapter 4

Experiment

In this section, we are going to introduce the experiment architecture designed to

evaluate 3N performance. With the performing of the experiment on multiple

kinds of platforms, the experiment result data will be listed to find out if the

implementation of 3N architecture is successful.

4.1. Experiment Architecture

Before performing experiments on complex network environment, a proper network

architecture should be settled. From the evaluation architecture of the example from

nnnSIM, a tree topology is proposed to suit with the network environment. Figure 9

shows the tree topology which contains one name producer also as the content producer,

router nodes in the middle and one moving mobile consumer roams from wireless

sector to sector.

This tree topology is designed to be the minimum standard structure to evaluate 3N

functions, which includes 3N naming, 3N packet transferring and 3N roaming scheme.

26

Fig.9. Basic Tree Topology

4.2. Experiment Environment

There are two main experiments performed in this research, lab experiment which uses

physical connections for the 3N architecture, and programmable virtualized network

environment experiment performed in VNode with the support of NTT.

4.2.1. Lab Experiment Configuration

For the setups of the lab experiment, we sign each 3N node with its own nnnSIM

environment in separated hosts. The hard line connections are provided by cables, while

27

the wireless connections are provided by 2.4GHz Wi-Fi access points (APs). Figure 10

shows the setup structure of the lab experiment, the consumer roams from AP to AP

during the experiment. However, there is a problem with the control of the roaming

process. While the evaluation is between the simulated scenario and implemented

scenario, it is hard to control the wireless handoff procedures to act the same as in the

simulator. Therefor we decided to perform experiment with two options, preset the

handoff timing, or use the nnnSIM to only emulate the handoff process area of the

experiment. Figure 11 shows the replacing part of the second option, the roaming

scenario contains four APs and one mobile consumer, the same as above. The four APs

are connected via FdNetDevice to the upper router nodes separately.

Table 5 shows the specs applied in the lab experiment.

TABLE.5. Roaming Schedule in Lab Experiment

VNode

DEN REN

100

Packet/s

1st 16s 19s

2nd 32s 35s

3rd 46s 49s

28

Fig.10. Ideal 3N Experiment Configuration

29

Fig.11. Emulated Roaming Process

4.2.2. VNode Experiment

For the experiments performed over VNode, there are actually two types of VNode,

VNode and VNode2. VNode is considered to have many servers separated in different

locations, while VNode2 could be recognized as a minor change version of VNode.

The experiment structure in VNode is similar to that in the lab experiment, shown by

Figure 10. Figure 12 shows the experiment structure in VNode2, in the scenario we

only applied 2 APs because the equipment storage was low, but still the experiment is

able to test the roaming process in 3N architecture.

Table 6 shows the specs of the experiment in VNode2.

30

TABLE.6. Roaming Schedule in VNode2

VNode2

DEN REN 100

Packets/s 45s 50s

Fig.12. Experiment Structure in VNode2

31

4.3. Experiment Results and Evaluation

Here we are going to show the experiment results. Figure 13 shows the throughput of

the lab experiment, we can see the roaming gap are quite big, this is because the wireless

device we applied in the experiment suffers long handoff due to its hardware limitation.

So we have to set the schedule long enough to cover the handoff procedure.

From the result, we can still tell that the throughput is quite smooth which means there

are not duplicated requests generated during the roaming process, which indicates the

success of performing a seamless roaming experiment.

Fig.13. Throughput of Lab Experiment

Figure 14 and Figure 15 show the compare result of the lab experiment and VNode2

experiment. From the compare we can see that even the content line is not smooth as

above, the interest line remains the same as requested, which indicates there are no loss

of data during the handoff. The burst of the content bitrate is because of the wireless

32

I/O performance and bandwidth limitation.

Fig.14. Throughput of Emulated Experiment with 1 switch

Fig.15. Throughput of VNode2 Experiment

33

Chapter 5

Conclusion

The work in my master thesis was to find a solution managed to implement 3N

architecture in complex and even deeply programmable network for evaluation. Aiming

the purpose of improving mobility in ICN structure network, in which routing strategy

used Face ID to forward content to the destination. With node naming function applied

in ICN, roaming consumers can retrieve their desired contents from previous request

from now on. With the powerful support of VNode testbed, it is very important and

convenient to implement 3N into complex network environment and improve the

performance of the system.

Future Work

In the future, along with the vision of future network and network appliances with ICN,

3N architecture is a promising method to improve the quality of mobile devices’ using

experiences. This function is playing a vital role in the field of live data stream service,

such as VoIP and HD Video Stream. For this purpose, implementations of 3N

architecture under more practical network environment are required in the short future.

34

References

[1] Jacobson, V. Networking named content. In Proceedings of the 5th International

Conference on Emerging Networking Experiments and Technologies (New York,

NY, USA, 2009), CoNEXT ’09, ACM, pp. 1–12.

[2] George Xylomenos, Christopher N. Ververidis, Vasilios A. Siris, Nikos Fotiou ;

Christos Tsilopoulos ; Xenofon Vasilakos ; Konstantinos V. Katsaros ; George C.

Polyzos, A Survey of Information-Centric Networking Research. IEEE

Communications Surveys & Tutorials (Volume:16 , Issue: 2 ) 2014 pp. 1024 –

1049.

[3] Gareth Tyson, Nishanth Sastry, Ruben Cuevas, Ivica Rimac, Andreas Mauthe. A

Survey of Mobility in Information-Centric Networks. Communications of the

ACM, Vol. 56 No. 12, 2013. Pages 90-98.

[4] NDN Project. [Online]. Available: https://named-data.net/project/

[5] Project ndnSIM. [Online]. Available: http://ndnsim.net/

[6] Project CCNxTM, PARC, 2013. [Online]. Available: http://www.ccnx.org, Sep.

2009.

[7] Jairo E. Lopez, Mohammed A., Li Zhu, Zheng Wen, and Sato Takuro, “Seamless

mobility in data aware networking,” in ITU Kaleidoscope 2015 - Trust in the

Information Society. 2015, ITU.

[8] ns-3 Developers, "ns-3," NS-3 Consortium, 6 2015. [Online]. Available:

https://www.nsnam.org/. [Accessed 25 9 2015].

[9] ndnSIM v1.0. [Online]. Available: http://ndnsim.net/1.0/

[10] K. Yamada; Y. Kanada; K. Amemiya; A. Nakao; Y. Saida, "VNode infrastructure

enhancement - Deeply programmable network virtalization," 2015 21st Asia-

Pacific Conference on Communications (APCC), pp. 244-249, 2015.

[11] FdNetDevice, Module of ns-3. [Online]. Available:

https://www.nsnam.org/docs/models/html/fd-net-device.html

35

Publication List

[1] Xin Qi, Lu Zhang, Zheng Wen, Takuro Sato. “3N-SIMULATOR PACKAGE

TRANSMISSION ON PHYSICAL NETWORK”. IEICE General Conference,

March 2016, Kyushu University

[2] Lu Zhang, Xin Qi, Di Zhang, Takuro Sato. “A Data Center Architecture Based on

CCN”. IEICE General Conference, March 2016, Kyushu University

[3] Xin Qi, Zheng Wen, Wataru Kameyama, Jiro Katto, Takuro Sato (Waseda

University), Yuki Minami, Kazuhisa Yamada (NTT). “Experimental Results of

Named Node Networking (3N) in VNode on Network Virtualization”. IEICE

Technical Committee on Network Virtualization (NV), The 18th domestic

conference, April 2016, Kikai Shinko Kaikan.