Algorithm for VOD Services using Multicast
Iksoo Kim and Yoseop Woo
Department of Information and Telecommunication Engineering University of Incheon
177 Namku Tohwa-dong Incheon, Korea
{iskim , yswooo}@incheon.ac.kr
Abstract – This paper presents algorithm for VOD
system using multicast delivery in Head-End-
Network (HNET). This is not conventional server-
based multicast but provides client-based on-
demand one. The scheduler generates immediately
multicast group address and port number when a
client requests a video, and then sends them to VOD
server and the client. And then VOD server
transmits requested video to multicast group and the
client joins the group automatically. The scheduler
assigns the same multicast group address when
other clients request an identical video within the
same scheduling duration. Otherwise, it assigns
another multicast group address to them. This paper
has been focus to implement VOD system with
JDK1.4 and JMF.
I. Introduction
The Multimedia information service is evolving to the
on-demand service, which can be provided to the
consumer whenever or wherever he or she wants it [1,
2]. Two of the best ways to improve performance of
VOD systems are multicast delivery technique and web
caching strategy using proxy [3, 4, 5, 6].
The multicast has been shown to be very effective in
reducing the server bandwidth. The conventional
multicast delivery systems announce service time for a
specific item in advance, and then clients access media
server at that time. And conventional VOD services
using multicast as a kind of periodic broadcast
periodically provide video service at prescheduled
service times. Consequently, these systems do not
support on-demand service for clients and they are
server-based multicast system. Thus they generate a
multicast group address and port number in prior to
clients’ request, and all clients who want to view the
whole video have to access VOD server at prescheduled
service time [7, 8, 9].
The proxy systems can improve network performance
and reduce the load of media server. It is a special kind
of server located between server and client. The conven-
tional caching techniques have unbalanced problems for
the load of proxies and the duplicated copy among proxy.
Thus the proxies storing popular items are faced with
heavier traffic than other proxies [4, 6]. Therefore,
proxy-caching systems have to solve imbalanced load of
proxies [10].
In this paper, the implemented VOD service using
multicast delivery technique can support on-demand
service. For providing on-demand service, scheduler on
server side assigns a multicast group address and port
number the instant that client requests a VOD service
and sends them to VOD server and client. Thus this
system has the characteristics of on-demand. The
scheduler aggregates clients’ request for an identical
video item within predefined scheduling duration (20
seconds) in order to maximize the efficiency of VOD
server and network resources. Therefore the clients who
request an identical video item within the same
scheduling duration receive the same multicast group
address and port number. We implement this system
with JDK1.4 and JMF (java media framework) [11]
The rest of the paper is as follows: Section 2
describes the structure and algorithm for VOD system
using client-based multicast delivery in detail, and
addresses the implementation of VOD systems
supporting multicast delivery with JDK 1.4 and JMF,
Section 3 shows the result of improvement of proposed
VOD system through simulation, and finally Section 4
concludes the paper.
II. Algorithm of VOD System using Multicast
Delivery
The VOD system using multicast delivery in this paper
consists of VOD server, scheduler on server side and a
number of clients connected to their ADSL/Cable
modem network as high-speed network, as shown in
Fig. 1 [10, 12].
The VOD server immediately performs streaming
service through Internet as a source device that provides
VOD service of a requested movie. It transmits video
streams requested to specific multicast groups that are
transmitted from scheduler.
The scheduler is a central control point that
aggregates service requests from clients and assigns
multicast group address and port number through
scheduling process. Then it sends multicast group
address and port number to VOD server and clients
requesting service. The scheduler assigns the same
multicast group address and port number to clients
when the thread for an identical video item requested
exists. Otherwise, it assigns another multicast group
address and port number to client. Thus the scheduler
creates different multicast address and port number to
clients who request an identical item after the each
grouping duration passed. Therefore this system may
generate a number of multicast group addresses and
port number for an identical video.
Head-end-Network(HNET) II
HEN a
VOD Server
HEN b
Internet
HEN c
Scheduler
Head-end-Network(HNET) I
Fig. 1 The structure and operation diagram for VOD System
The VOD server transmits video streams as soon as
receiving multicast address and port number from
scheduler and clients connected to HENs join a
multicast group automatically after receiving them. The
VOD server transmits advertisement contents during
grouping time. There are two reasons for sending it; to
remove the latency for service from VOD server and to
decrease the service charge. Although the clients who
request an identical item during a same grouping
duration are assigned the same multicast group, the time
of viewing the advertisement is different for each client.
The first client who requests service views it for entire
grouping durations, but other clients view it for less
than grouping duration according to their request instant
within a grouping duration. But the instance of real
video viewing for the clients that join same multicast
group address and port number is same.
The scheduler divides into two parts, main scheduler
and socket thread. The main scheduler using server
socket receives clients’ request, generates multicast
groups according to scheduling duration to each
requested video item and manages them. The socket
thread maintains connection of client. Fig. 2 and Fig. 3
show overall service procedure of proposed system and
scheduler procedure, respectively
Step 1: Client accesses scheduler for VOD
service request: inputs requested video
item’s ID
Step 2: The scheduler investigates whether the
scheduling session for requested video
ID exists or not.
If it exists go to step 4.
Step 3: The scheduler performs grouping and
generates multicast group address and
port number during grouping duration.
Step 4: The scheduler sends multicast group
address and port number to VOD
server and clients who request an
identical video item.
G:\New Paper\호주\Atnac2003\Iksoo Kim.doc
Step 5: Server transmits requested video to a
received multicast group.
The clients join the received multicast
group and port number.
Step 6: Java Media Player play outs video.
Step 7: Complete video service.
Fig. 2 Overall VOD service and Join procedure
Step 1: Initiate scheduler
Step 2: The scheduler waits until the clients request
VOD service.
Step 3: Scheduler thread wakes up when the client
requests service.
It receives video item ID from client.
Step 4: The scheduler investigates whether the
scheduling session for requested video ID
exists or not.
If it exists, then go to step 6.
Step 5: The scheduler performs grouping and
generates multicast group address and port
number during grouping duration.
Step 6: The scheduler sends multicast group
address and port number to VOD server
and clients who request an identical video
item.
Step 7: Completion scheduling
Fig. 3 Scheduler on server side procedure
Fig. 4 and 5 show server and client procedure,
respectively. The VOD server generates Datagram
socket and inserts the multicast group address and port
number within the datagram packet, and transmits video
streams to clients. It divides into two parts, the one is
Datagram socket as a maintaining indirect connection
with clients and the other is data transmitting for
sending requested video item.
Step 1: The scheduler calls VOD server: VOD
server starts.
Step 2: VOD server receives multicast group
address and port number from
scheduler.
Step 3: VOD server investigates compression
format and the rate of transmission.
Step 4: VOD server transmits video streams
with received multicast address and
port number.
Step 5: VOD server completes transmission of
video streams; Completion server
program
Fig. 4 VOD server procedure
Step1: Client generates socket and prepares
connection to scheduler.
Step 2: Client requests video items ID to
scheduler after connection.
Step 3: Client receives multicast group address
and port number from scheduler.
Step 4: Client closes socket for scheduling
17/10/2003
 Step 5: Client joins multicast group address in
order to receive transmitted video
streams from VOD server.
Step 6: Java Media Player plays out video
streams.
Step 7: Java Media Player stops play-out.
Fig. 5 Client procedure
Fig. 6 shows screen shot of scheduler frame. As
shown in Fig. 6, the scheduler on server side indicates
grouping requested video items list on window of
grouping video item during a grouping duration.
The grouping item lists are added requested video
item IDs when grouping item field within
CommonData class becomes set. Also, the scheduler
indicates generated multicast group addresses, port
numbers, requested video IDs, connected clients’ IPs
which request service and requested video items on
window of User.
Fig. 6. The scheduler application on server side
Fig. 7 and 8 show the results of execution client
program and GUI of client application, respectively. As
shown in Fig. 7, the execution of client program
performs on command line through ClientFrame. The
result of execution is the window of client GUI (Fig. 8).
And then client writes server IP or DNS name of
scheduler on Client Frame (Fig. 8), well-known port
number and video item number on corresponding
window (Fig. 8). The port number 7777 is a port that
the function of ServerSocket( ) uses. Then the result of
scheduling by the execution of thread is transmitted to
the client from scheduler. The transmitted multicast
group address (232.0.175.217) and port number (4096)
are written on the bottom of Fig. 7 and in the TextArea
of Fig. 8. And then, the client automatically joins with
G:\New Paper\호주\Atnac2003\Iksoo Kim.doc
232.0.175.217 and 4096(port number) using MClient
class and SimplePlayer class.
The MClient class generates socket to communicate
with scheduler, and it outputs multicast group address
and port number on client’s window. Next, MClient
class closes socket and attempts to join a multicast
group address and port number, and then generates
SimplePlayer class.
Fig. 7 The execution client program
Fig. 8 GUI of client application
Fig. 9 shows the results of execution client’s applet
application. The client writes server IP or DNS name of
scheduler on Client Frame (Fig. 8), well-known port
number and video item number on corresponding
window. The port number 7777 is a port that the
function of ServerSocket( ) uses. Then the result of
scheduling by the execution of thread is transmitted to
the client from scheduler. The transmitted multicast
group address (230.0. 127.240) and port number (4044)
are written on the bottom of Fig. 9. And then, the client
automatically joins with 230.0.127.240 and 4044(port
number).
The SimplePlayer class drives Multicast Media
Player using JMF (java media framework) and receives
video streams transmitted from VOD server through
RTP session with a multicast group address. The
Multicast Media Player performs rendering of received
video streams. The result of execution of Fig. 9 (click
request button) is shown in Fig. 10. Thus, Multicast
Media Player with controller plays out video streams
transmitting from server
17/10/2003

Fig. 9 Applet window for client execution
Fig. 10 Multicast Media Player using JMF
III. Simulation and Results
This section performs simulation and analyzes on the
results of performance of media server using multicast
delivery. The request patterns depend on the popularity
of the video serviced from media server. We use Zipf-
like distribution for the probability of choosing i-th
popular video among N multimedia items provided by
media server. The probability that a specified
multimedia is chosen with ith frequency is Z/i, where
Z=1/(1+1/2+1/3+ . . . +1/N). Using the Zipf distribution,
if the overall clients’ service request rate to the media
server through scheduler is λ, the service requesting
rate of the multimedia item for ith frequency is λi=λρ
i(where, ρi =Z/i)). Its rate based on popularity is used
to determine the traffic for each multimedia item. The
most popular multimedia item (i=1) must have higher
weighted value than others because the request for
popular multimedia item is more frequent than that for
unpopular ones. We use ρi as a weighting parameter
G:\New Paper\호주\Atnac2003\Iksoo Kim.doc
for selection of each multimedia item in simulation.
Consequently, we perform simulation such that the
more popular multimedia items, the higher probability
of request for the multimedia item [13, 14].
In the simulation, the number of multimedia item
provided by the media server is N=100 and the length
of each multimedia is 1 to 30 minutes, and the service
request rate(λ) per head-end network(LAN) is 30 to
100 per minute. The service request rate λ follows
Poisson distribution and the expression is as follows
[15],
fx(X = k) = lim ( n
k
) pk (1 - p)n-k = λk e-λ / k! (1)
n∞
Fig. 11 shows the number of required channel for
service as the service request rate λ = 30 to 100 vary.
The performance of proposed media system decreases
the number of channel by 32.1%, 46.7% and 58.4%
when the service request rate λ = 30, 50 and 100,
respectively.
120
100 The Bandwidth of
Unicast is 100%
80 67.9%
53.3%
60 41.6%
40
20
0
0 30 50 100
λ : Request
Fig. 11 The result of simulation for the load of server
IV. Conclusion
The proposed system implements VOD service using
client-based multicast delivery system not server-based
one. The scheduler immediately generates multicast
group address and port number according to clients’
request. The VOD server starts sending requested video
to group and the clients automatically join that group
and play out receiving video streams. This VOD system
shows that the number of channel decreases by 32.1%,
46.7% and 58.4% in comparison of that using unicast
when service requesting rates are 30, 50 and 100,
respectively. We believe that the range of application
will be extended when the proposed system supports
multicast tunneling on Internet and VCR operation.
Acknowledgement
This work was supported (in part) by the Korea Science
and Engineering Foundation (KOSEF) through the
multimedia Research
Reference
17/10/2003
1. K.Almeroth & M.Ammar, "An Alterative Paradigm for
Scalable On-Demand Application: Evaluating and
Deploying the Interactive Multimedia Jukebox", IEEE
Trans. on Knowledge and Data Engineering Special Issue
on Web Technology, April 1999
2. D.Sitaram and A.Dan, “Multimedia Servers”, Morgan
Kaufmann Publishers 2000
3. T.D.C. Little and D. Venkatesh, "Prospects for Interactive
Video-On-Demand" In MCL Technical Report 02-15-1994
4. Carey Williamson, “On Filter Effects in Web Caching
Hierarchies,” ACM Transactions on Internet Technology,
Vol. 2, No. 1, pp. 47-77, February 2002
5. Backhyun Kim, Seungchan Moon, Iksoo Kim & Yoseop
Woo, "A Buffering Algorithm for Providing Truly
Interactive Video-on-Demand Service," Proc. of Int'l
Conf. on PDPTA99, pp211-217, Las Vegas, 1999
6. R. Rajaie, M. Handley, H. Yu and D. Estrin, “Proxy
caching mechanism for multimedia playback streams in
Internet,” in Fourth International WWW Caching
Workshop, Mar. 1999
7. Yen-Jen Lee, David H.C.Du and Weihsiu Ma, "SESAME: A
Scalable and Extensible Architecture for Multimedia
Entertainment", In IEEE, COMPSAC, 1996.
8. K.C.Almeroth and M.H.Ammar,"The Use of Multicast
Delivery to Provide a Scalable and Interactive Video-On-
Demand Service," IEEE J. Selected Areas in Comm., Vol.14,
No 6, Aug. 1996, pp.1110-1122
9. J.Pasquale, G.Polyzos and G.Xylomenos, “The Multimedia
Multicasting Problem,” ACM Multimedia Systems, vol.6,
no.1, 1998
10. Iksoo Kim, Taejun Hwang, Youngjune Kim & Yoseop
Woo, " New Distributive Web caching Technique for VOD
Service," ITCOM2002 Conf. Boston U.S.A, 2002
11. http://java.sun.com/products/java-media/jmf/
12. Iksoo Kim, Backhyun Kim , Yoseop Woo , Taejun
Hwang , Seokhoon Kang,”VOD Service using Web-
Caching Technique on the Head-End-Network,”
ICCSA2003 Springer-Verlag LNCS2668
13. P. Cao, L. Fan and G. Philips, “Web Caching and Zipf-
like Distributions: Evidence and Implications,” IEEE
Infocom 1999
14. D.Eager, M.Vernon & J.Zahorjan, " Optimal and Efficient
Merging Schedules for Video-on-demand Servers", Proc.
ACM Multimedia 99, ACM 1999
15. H. M. Taylor and S. Karlin, “An Introduction to
Stochastic Modeling,” Academic Press 1994

G:\New Paper\호주\Atnac2003\Iksoo Kim.doc 17/10/2003