On the Packet Similarity of Real-time Streaming Services
Takumi MIYOSHI and Kenji SEKIYA
Graduate School of Engineering
Shibaura Institute of Technology
307 Fukasaku, Minuma-ku, Saitama-shi, Saitama, 337-8570 Japan
Email: {miyoshi,m109053}@shibaura-it.ac.jp
Abstract—Multimedia streaming has recently become a dom-
inant form of Internet traffic. Even in a live streaming en-
vironment, however, each end user independently connects to
the server by unicast communication. It is thus essential to
consider how to provide such services to many users while
maintaining efficiency and scalability. To explore the potential of
introducing multicast-like transmission mechanisms, we analyze
the similarity in streamed data packets that are simultaneously
transmitted to several end hosts in real-time video/audio services.
The results show a significant level of similarity among the
packets and, therefore, that multicast-like mechanisms could
decrease streaming traffic.
Index Terms—traffic analysis, packet similarity, streaming
service, IPTV, Internet radio
I. INTRODUCTION
With the popularization of the Internet, video/audio stream-
ing services have recently become a dominant form of traffic.
For multimedia live streaming, in which the same encoded
data are transmitted simultaneously to several receivers, mul-
ticasting is the preferred technique to decrease network traffic
[1]. However, unicast-based streaming services are currently
used in the majority of applications because they allow users
to enjoy video/audio without installing additional applications.
Thus the multicast mechanism is not applicable in the current
live streaming services.
On the other hand, alternative multicast-like mechanisms
have been proposed to decrease unicast traffic efficiently by
using multicast-like transmission techniques. Since they work
on the network side, neither the servers nor the end users need
to notice them; the traffic will simply be automatically sup-
pressed when the packets have the same or similar payloads.
In this paper, we analyze the packet similarity in real-time
video/audio streaming services. When several hosts receive the
streaming data, all the packets in each host are recorded using
a packet sniffer application. The packets are then compared
for whether they have the same or similar payloads.
II. MULTICAST-LIKE TRANSMISSION MECHANISMS
Explicit multicast (Xcast) is a type of application-layer mul-
ticast (ALM) [2]. Although Xcast uses only unicast packets,
the packets can be transmitted to multiple destinations using a
multi-hop mechanism on the application layer. Many extended
versions of the Xcast protocol have been developed. In [3], Iu
proposed an interesting mechanism that compresses multiple
unicast packets into an Xcast packet at Xcast-aware routers on
the way to their destinations. The Xcast packet is decoded to
the original unicast packets at the designated routers.
Striegel proposed a stealth multicast feature in [4]. Routers
equipped with this function encapsulate multiple unicast pack-
ets with the same data payload into a multicast or ALM
packet. End hosts do not need to recognize the stealth multicast
protocol.
In [5], [6], the authors also proposed and implemented
μSEcast. Multiple unicast packets that have the same or similar
data payloads are compressed at a μSEcast-aware router into
a single packet with multiple destinations. Since the μSEcast
protocol uses only unicast packet formats, it is completely
transparent to the traditional Internet protocol.
The three mechanisms above are very similar to one another
and are shown in Fig. 1. Since multiple unicast packets are
compressed into a single packet, the mechanisms provide more
efficient transmission in the network. The condition for com-
pression is that the unicast packets going into the router have
the same or similar data payloads. Therefore, it is important to
check whether streaming services simultaneously transmit the
same or similar packets to multiple end users. If so, then the
network traffic can be drastically reduced, especially in real-
time video/audio streaming services, by using multicast-like
transmission mechanisms. This is the main motivation for our
analysis.
III. HOW TO ANALYZE PACKET SIMILARITY
We focus on the following real-time video/audio streaming
services: (1) Nico Nico Live [7], (2) Livedoor’s net radio Net-
radi [8], (3) Radio Nikkei [9], and (4) USTREAM [10]. (1)
and (4), the live streaming video services in Japan and the
U.S., can be received on web browsers with the Adobe Flash
plug-in. (2) and (3) are network radio services that transmit
streaming audio data for Windows Media Player. We have
prepared two end hosts connected to the Internet, as shown in
Fig. 2. We view or listen to the same streaming data on these
Receiver 1
Receiver 2
Receiver 3
Transmission Overlay
123
3
2
1
Multicast-based transmission
(native multicast, ALM, tunneling)
Unicast packets
M
Fig. 1. Multicast-like transmission mechanism.
2009 International Conference on Intelligent Networking and Collaborative Systems
978-0-7695-3858-7/09 $26.00 © 2009 IEEE
DOI 10.1109/INCOS.2009.47
265