2022 5th International Conference on Information and Communications Technology (ICOIACT)

Design and Implementation Named Data

Networking-Based Video Streaming System
Ade Aditya Ramadha Leanna Vidya Yovita Tody Ariefianto Wibowo
2022 5th International Conference on Information and Communications Technology (ICOIACT) | 978-1-6654-5140-6/22/$31.00 ©2022 IEEE | DOI: 10.1109/ICOIACT55506.2022.9972215

Fakultas Teknik Elektro Fakultas Teknik Elektro Fakultas Teknik Elektro

Telkom University Telkom University Telkom University
Bandung, Indonesia Bandung, Indonesia Bandung, Indonesia
[email protected] [email protected] [email protected]

Abstract—Named Data Networking (NDN) is a future internet optimized to overcome this problem. With the advantage of
architecture with the method of giving a name to data and has the new networking architecture NDN and based on current
in-network content caching abilities. With this ability, users can user behaviour, NDN will fit more on video streaming systems
get their requested data faster. As users prioritize content or
data more than the origin of the data, NDN can give more than the old IP network architecture.
advantages than IP does. Video streaming is a technology that To improve video transmission on a video streaming system,
enables users to view live videos or pre-recorded videos without we developed a Named Data Networking (NDN)-based Video
having to download them. The development of video streaming Streaming System. This paper also compares IP-based video
services and access to video streaming services continues to
streaming systems and our implementation of an NDN-based
increase yearly. To overcome this problem, optimization of
video applications and video transmission need to be done. In video streaming system. The NDN-based video streaming
this paper, we propose a Video Streaming System based on system we propose uses the hls.js library and is implemented
future internet architecture Named Data Networking (NDN). on Packet Network Emulator Tool Lab (PNETLab).
This paper also compares IP-based video streaming systems
and our implementation of an NDN-based video streaming II. BACKGROUND AND R ELATED W ORK
system. Our NDN-based video streaming system can provide
some improvement on video streaming applications over the A. Video Streaming
existing IP-based video streaming system. As a result, our NDN-
based video streaming implementation shows that performance Video Streaming is a technology that can play audio
on second access improved because of the in-caching network or video files directly or pre-record without downloading
feature on NDN. the video first. At the beginning of its development, video
Index Terms—Video Streaming, Named Data Networking,
streaming had several protocols, such as Transmission Control
Video Streaming over NDN.
Protocol / User Datagram Protocol (TCP / UDP) and Real-
time Transport Protocol / Real-time Transport Control Protocol
I. I NTRODUCTION
(RTP / RTCP). TCP and RTCP were used for data trans-
The current internet network is based on host-to-host com- mission/control signalling. In contrast, UDP and RTP send
munication. This communication has been used since 1960 audio/video media. At that time, UDP was preferred. It had
and 1970 and aims to solve the problem of resource sharing. advantages such as faster delivery because it did not require a
This communication is no longer relevant to the current state of three-way handshake process like TCP [4] [5].
internet users who prioritize the content of data rather than the In addition to only having a few protocols for video stream-
origin of the data. Therefore, Van Jacobson proposed a new ing, at the beginning of its development, it also still had a fixed
network architecture, namely Named Data Network (NDN) data bit rate. Suppose the network performance decreases, the
[1]. time required to transmit data increases, so the buffering time
NDN network architecture assigns a name or address to a is longer and lowers the quality of streaming video. Therefore
data and adds caching of content in the network. With the we need a method that can adapt the data bit rate to reduce
addition of this caching, users can access content faster and buffering time when network performance decreases.
can be more easily accessed. In addition, users do not need to Several technologies implement this method using HTTP
know where the source of the content is because users only protocols such as HTTP Live Streaming (HLS) and Dynamic
need to send data requests without the need to include the Adaptive Streaming Over HTTP (DASH). HLS splits a video
address of the data owner [2] file into smaller parts with different data bit rates and qualities.
The demand for video content is constantly increasing in The input video/audio file will be served by the media encoder
today’s network and will become a problem in future networks. and stream segmenter to generate an extended M3U playlist
Cisco forecasts that in 2022 in the future, nearly 79% of file containing some Uniform Resource Identifiers (URI) of the
data traffic generated by mobile devices is video data traffic media segment of a particular video and the media segment,
[3]. The application video and video transmission must be which is the result of splitting the original video/audio file

Authorized licensed use limited to: Institut Teknologi Bandung. Downloaded on May 31,2023 at 06:59:36 UTC from IEEE Xplore. Restrictions apply.
978-1-6654-5140-6/22/$31.00 ©2022 IEEE 66
 2022 5th International Conference on Information and Communications Technology (ICOIACT)

Fig. 2: NDN Packet Structure.

Fig. 1: MPEG2-TS Packet Structure.

C. Named Data Networking

[4] [6]. Therefore, our system uses the HLS protocol and use Named Data Networking is a network architecture with a
hls.js library. method of naming data. This method allows the NDN client
to get a data packet by name. In addition, this architecture
can cache content in the network by adding a cache on
B. Video Chunking
the router. Network caching capability allows users to get
Generally speaking, video chunking on IP-based Video the requested data faster. Data requested by sending interest
Streaming and NDN-based video streaming share the same packet, the producer will send data using packet data. With this
principle. The actual video is cut into several segments and mechanism, it replaces request-response on IP architecture. To
labelled sequentially. Every segment will be encapsulated track pending interest packets, the NDN router has a record in
based on the protocol it uses. We will discuss encapsulated its Pending Interest Table (PIT). If any packet data match the
packets on NDN-based video streaming and IP-based video interest, the router will send the packet data according to PIT.
streaming. Another part of the NDN router is the Content Store (CS),
1) Encapsulated packets on IP-based video streaming: CS is used to store packet data. With CS, the NDN router
The content will be delivered using MPEG2-TS format in an can have a copy of the data packet, and thus if another client
IP-Based video streaming system. As shown in Fig. 1, The on the same local network requests the same data, the NDN
full video will be split into several segments according to router can give local data in CS and improve performance
the duration by hls protocol. The segment containing data on the network. With fundamental change and many features
like video or audio is called Elementary Stream (ES) on it provides, NDN can replace current network IP and give a
the MPEG2-TS system. These Elementary Streams are then more improved performance to content distribution systems
packetized to form Packetized Elementary Stream (PES) layer, like video streaming systems [2] [7].
and every PES is attached with an identifier called PID. The D. Related Work
PES packet will be attached with a TS header on the Transport
Several studies related to ndn-based video streaming have
Stream Level. On the Transport Stream layer, the overall
been conducted. Research [8] implements live streaming
packet will be fixed into 188 bytes with a 4 bytes header
(NDNLive) and video streaming (NDNTube) through the NDN
and 184 bytes max data. If the data surpass 184 bytes, the
network, both applications developed on top of the Consumer
PES packet will split into several TS level packet, and vice
/ Producer API. Research [9] implementing Proactive Content
versa; if the data is lower than 184 bytes, the data on Transport
Caching using dash-ndn-js in a train. Research [10] imple-
Stream Level will be stuffed. All packets will be sent using
menting streaming scenarios using the mobile DASH library
TCP protocol and using IP.
on the NDN architecture. Research [11] implements video
2) Encapsulated packets on NDN-based video streaming: streaming over a testbed NDN network connected to multiple
As shown in Fig. 2, The content will be sent using NDN Packet routers in a university or organization. This research uses the
in an NDN-based viddeo streaming system. Like an IP-based hls protocol and shaka-player as the client’s web video player
video streaming system, the entire video will be split into library. The drawback of research [8], [9], [10], and [11] is the
several segments according to the duration. Each segment will lack of support for the new NDN packet format specification.
be split into several chunks to match the 8 KB data packet on In research [11], the goal is to reduce configuration on the
NDN. To send these chunks, the data packet will be generated client-side, so in their system, the client has to load the
by attaching an NDN Header and Signature of the data. NDN required library from the internet leading to lower performance
Header will contain the meta info like Freshness period to on video startup delay.
determine when the data is considered stale and Final Block
Id to determine the last chunk ID. The Signature part will III. NAMED DATA N ETWORKING -BASED V IDEO
contain Signature Info and Signature Value to ensure the data S TREAMING S YSTEM
received on the client is the same as the data sent from the This paper uses a network emulator as the base for the
server. implementation and evaluation of NDN based video streaming

Authorized licensed use limited to: Institut Teknologi Bandung. Downloaded on May 31,2023 at 06:59:36 UTC from IEEE Xplore. Restrictions apply.
67
 2022 5th International Conference on Information and Communications Technology (ICOIACT)

Fig. 3: Research Flow Diagram.

Fig. 5: Topology Used on This Paper.

press as a framework for web servers and handles some routes.

Node.js is used because they are easy to use, deploy, and fast.
Route / will respond with index.html contains the main page
for the video streaming web page. Route /hls will respond with
the hls.js library for activating the video streaming webpage.
Fig. 4: Proposed Video Streaming System Method. Route /segment-list will request the playlists file to the server
over the NDN network using ndncat catch and return it to
the client’s web browser. Route /segment/:segments, similar to
system. The network Emulator used in this paper is Packet route /segment-list, request the segmented video to the server
Network Emulator Tool Lab (PNETLab). The research flow over the NDN network using ndncat catch and return the file
diagram is shown in Fig. 3. to the client’s web browser.
Our video streaming system uses hls. Fig. 4 shows our Besides Node.js, the client will also run a web browser
video streaming system. On the server/producer side, the to access the video streaming server and play the video
server provides a playlist file and video files segmented into streaming. NFD also plays a significant role in processing the
several segments called representation and serves it as an NDN packet like sending the interest packet, reading packet
NDN packet. On the client side, it uses the hls.js library, data, and even storing the data in the content store.
node.js as a web server, and ndncat to retrieve segments from
the server/producer. Several nodes activate the NDN packet C. Network
forwarder function between client and server or producer and Our NDN-based video streaming system is implemented on
act as NDN routers. a network emulator called Packet Network Emulator Tool Lab
(PNET-Lab). PNET-Lab uses KVM as the base of networking
A. Server Side node emulation. PNET-Lab also supports Network Element
On the server in Fig. 4, the video representation is served based on Docker, IOL, Dynamips, and QEMU. With this
as an NDN packet. For simplicity, in this paper, we use feature, we can emulate almost every device in existence.
the hls protocol with a 1080p quality video and 6 seconds Because using KVM as the base of emulation, PNET-Lab can
on each segment. We use 6 seconds as recommended by perform close to actual hardware, thus reproducing data close
apple. As shown in Fig. 4, the segmented video is served to real hardware. NDN router along the network uses NDN
using ndncat put. The client can request each chunk of seg- Forwarding Daemon (NFD) as NDN packet forwarder. NFD
mented video using an interest packet. The naming convention also activates the function of the NDN router, including the
on ndncat put follows /ndn/video/(filename)/(version)/(chunk- Content Store. We use NFD because (1) It is supported by a
number). When the server receives an interest packet, the prominent member of the NDN community, and (2) Always
server will send the corresponding segment according to updates to the latest NDN packet format specification.
the interest packet. The Server also runs NDN Forwarding
Daemon (NFD) to activate the NDN function and can process IV. E VALUATION
NDN packets. We conduct several experiments with several parameters
to get results in different conditions for the evaluation. On
B. Client Side client side, we measure the performance using QoE parameters
As shown in Fig. 4, the client runs several applications. such as video startup delay and rebuffering. On the Server
There are a web browser, Node.js, and NFD. Node.js uses ex- side, we measure the CPU usage to see how each request

Authorized licensed use limited to: Institut Teknologi Bandung. Downloaded on May 31,2023 at 06:59:36 UTC from IEEE Xplore. Restrictions apply.
68
 2022 5th International Conference on Information and Communications Technology (ICOIACT)

Fig. 6: Video Startup Delay Result. Fig. 8: Throughput Result.

on first access to 7.5 s-8.3 s on second access and shows up

to 20% improvement. Whereas the IP-based system does not
show any improvement on the second access from 1.3 s-1.8 s
on first access to 1 s1.7 s on second access. Even though the
value of an IP-based system has better performance than the
NDN-based system, the NDN-based system can improve the
second access because of the in-network caching function on
NDN.
According to [12], the average user preferred startup delay
of under 2 seconds and the increase in video streaming delay
causes client interest in watching videos to decrease by 5.8%
every second. Thus NDN-based video streaming system still
Fig. 7: Round Trip Time Result. needs some improvement. The startup delay depends on the
topology it uses on an NDN-based video streaming system.
In the best-case scenario, the value of startup delay is under
affects server performance. On the network side, we use RTT 7 seconds, including time to load the necessary library and 6
and Throughput. We also compare our system with a video seconds of buffered video. None of either system, the client
streaming system based on IP. experienced rebuffering.
The experiment was conducted using four different scenar- 2) Network Side: On the Network side, we measure and
ios: (1) 1 Hop: Server only one hop apart from the client, (2) compare RTT and Throughput for NDN and IP-based video
5 Hop: Server is five hops apart from the client, (3) 10 Hop: streaming. The pattern from the client side is still shown on the
Server is ten hops apart from the client, (4) 5 client: Server is network side. The RTT and Throughput of NDN-based system
ten hops apart from the five clients. Scenarios 1,2 and 3 use shown in Fig. 7 and 8 show improvement on second access
two clients, and both clients will access the video streaming from 4.7 s-6.8 s on first access to 4.7 s-4.4 s on second access
alternately, and these scenarios will measure performance on and shows up to 35% improvement on RTT, and 0.5 MBps-
the client and network sides. Scenario 4 uses five clients to 0.8 MBps on first access to 0.82 MBps-0.87 MBps on second
access the video streaming alternately, and this scenario will access and shows up to 34% improvement on throughput. The
measure performance on the server side. All scenarios share RTT and Throughput of IP-based system shows a variety of
the same primary topology as shown in Fig. 5 improvements from 12% up to 24% improvement on RTT
and 11% up to 23% throughput improvement. Even though
A. Result the IP-based system gives more performance than NDN-
1) Client Side: On the client side, there are two metrics to based system, the NDN-based system improve performance
measure QoE. The first metric is video startup delay, which at second access because of the helpful in-network caching
measure the time to wait until the video starts playing. The function on NDN.
second is the number of rebuffering metrics to measure how We use NFD as an NDN forwarder. The only drawback
often video playback stopped/stalled. of using NFD as an NDN Packet forwarder is the lack
On the client side, we measure and compare video startup of multi-thread forwarding, thus affecting the NDN network
delay and rebuffering for NDN-based and IP-based video performance. The NDN-based video streaming system needs
streaming. As shown in Fig. 6, the NDN-based system has some improvement on the NDN forwarder or using another
improved performance on the second access from 7.9 s-10.3 s NDN forwarder that supports multi-threading. With the sup-

Authorized licensed use limited to: Institut Teknologi Bandung. Downloaded on May 31,2023 at 06:59:36 UTC from IEEE Xplore. Restrictions apply.
69
 2022 5th International Conference on Information and Communications Technology (ICOIACT)

Fig. 9: CPU Usage on both Systems.

port of multi-threading, more packets can be forwarded si- VI. F UTURE R ESEARCH
multaneously. This solution will increase the NDN Network This paper uses NFD as an NDN forwarder with lower
performance, such as RTT and Throughput. performance on forwarding than other NDN forwarders like
3) Server Side: On the Server side, we measure the CPU NDN-DPDK, MW-NFD, and YaNFD. We suggest trying this
Usage and compare the NDN-based system and the IP-based system with other forwarders to improve network side per-
system. Based on Fig. 9, the NDN-based system shows high formance. An improvement on video streaming applications
CPU usage only on the first access with an average CPU usage such as using newer protocol Quick UDP Internet Connections
of 12.18%, on the second access up to the fifth access shows (QUIC) and HTTP/3; integrating players on video streaming
low CPU usage with an average CPU usage from 1.34% up websites with NDN, so the video streaming system does
to 1.85%. On the IP-based system shown in Fig. 9, the first not need conversion from HTTP to NDN or vice versa.
access up to the fifth access shows the same CPU usage with Implementing NDN architecture on third layer of OSI will
no drop in CPU usage with average CPU usage on 0.12% up also improve forwarding on the NDN network.
to 0.22%. Even though NDN CPU usage is higher than the
IP-based system, NDN still gives an advantage on the second R EFERENCES
access and fifth access. This value shows in-network cache [1] V. Jacobson, D. K. Smetters, J. D. Thornton, M. F. Plass, N. H. Briggs,
function on NDN not only gives more improved quality but and R. L. Braynard, “Networking named content,” 2009.
[2] L. Zhang, A. Afanasyev, J. Burke, V. Jacobson, k. claffy, P. Crowley,
also spreads the CPU usage on the server side to nearby NDN C. Papadopoulos, L. Wang, and B. Zhang, “Named data networking,”
routers. SIGCOMM Comput. Commun. Rev., vol. 44, no. 3, p. 66–73, Jul. 2014.
The in-network caching will lower CPU usage on the server [3] Cisco visual networking index: Global mobile data
traffic forecast update, 2017–2022. [Online]. Available:
and spread the load to a nearby router on NDN Network. This https://s3.amazonaws.com/media.mediapost.com/uploads/CiscoForecast.pdf
feature gives an advantage to the server to process another [4] I. M. Widyantara, M. Bazly, and N. Er, “Mekanisme segmentasi laju bit
task that is more important than a repetitive process from many pada dynamic adaptive streaming over http (dash) untuk aplikasi video
streaming,” Majalah Ilmiah Teknologi Elektro (Journal of Electrical
users requesting the same content or video. The advantage will Technology), vol. 14, pp. 18–21, 12 2015.
increase if there are more NDN routers on the topology, so the [5] X. Ma and J. Gao, “The comparison and analysis of the streaming media
load will be evenly spread and perform better. The increased transport protocol in the transmission system,” 2012.
[6] B. Choi, G. Yang, and J. Kim, “Implementation of hls protocol for an
NDN routers also lead to lower resources used on every router. ip camera,” pp. 111–114, 04 2014.
[7] A. Afanasyev, J. Burke, T. Refaei, L. Wang, B. Zhang, and L. Zhang,
V. C ONCLUSION “A brief introduction to named data networking,” pp. 1–6, 10 2018.
[8] L. Wang, I. Moiseenko, and L. Zhang, “Ndnlive and ndntube : Live and
From the experiments that have been done, NDN-based prerecorded video streaming over ndn,” 2015.
Video Streaming could give more quality on second access [9] T. Muto, K. Kanai, and J. Katto, “Implementation evaluation of proactive
content caching using dash-ndn-js,” pp. 2239–2244, 06 2015.
thanks to its in-caching network feature. Even though the [10] D. Stohr, F. Beji, R. Dwarakanath, R. Steinmetz, and W. Effelsberg,
current NDN-based video streaming system still needs im- “Mobile ndn-based dynamic adaptive streaming over http,” p. 3, 01 2016.
provement on NDN architecture and improvement on NDN [11] C. Ghasemi, H. Yousefi, and B. Zhang, “Internet-scale video streaming
over ndn,” 2020.
application to give more or the same performance compared [12] S. Krishnan and R. Sitaraman, “Video stream quality impacts viewer
to the IP-based video streaming system. In conclusion, NDN behavior: Inferring causality using quasi-experimental designs,” Net-
can give more advantages than IP-based video streaming. Thus working, IEEE/ACM Transactions on, vol. 21, pp. 2001–2014, 12 2013.
an improved NDN-based video streaming system will replace
the current IP-based video streaming system.

Authorized licensed use limited to: Institut Teknologi Bandung. Downloaded on May 31,2023 at 06:59:36 UTC from IEEE Xplore. Restrictions apply.
70