Open-Loop Algorithms - Scalable Continuous Media Streaming Systems - page 321

0.6

CB, m= 1
CB, m= 2
CB, m= 4
CB, m= 16
LB

1 PHB, m= 1 and m =16, are

close to the LB of 37% of
video size

Sufficient buffer
requirement for the LB as
defined in Section 18.3

PHB, m= 1
PHB, m= 16

1

t

PB

GDB3

0.3

SB

SDB

SB
GDB3
SDB
PHB
CB
PB
LB

: Skyscraper Broadcasting
: Greedy Disk-Conserving Broadcasting
: Staircase Data Broadcasting
: Poly-harmonic Broadcasting
: Consonant Broadcasting
: Pagoda Broadcasting
: Lower Bound

0

2 b 4 b 6 b 8 b 10 b

Network bandwidth (multiples of video bit-rate b )

Figure 18.5 Client buffer to video size ratio versus network bandwidth

system complexity in terms of the number of channels required for broadcasting the media
segments.
For example, given a client access bandwidth constraint of 2 b and network bandwidth of
8.98 b , CB with m 4 can achieve a start-up latency of only 5.76 seconds but this requires

5,000 network multicast channels. In networks with limited number of multicast channels
(e.g., group addresses in IP multicast), this requirement can become a significant bottleneck.
To tackle this problem, we present in the following chapter a Grouped Consonant Broadcasting
(GCB) scheme to dramatically reduce the number of network channels required, with a small
trade-off in performance.

=

18.6.1 Bandwidth Partitioning and Reception Schedule
Type-I channels in GCB are the same as the original CB as defined in equations (18.5) and
(18.6). The difference is in the design of the Type-II channels. In CB, reception of Type-II
channels in the same group begins at the same time but ends at different times due to the just-
in-time scheduling principle. While this technique can reduce the bandwidth requirement, it
also requires the use of a separate network transmission channel (e.g., an IP multicast address)
for each of the Type-II channels.
To reduce the number of channels needed, we modify CB such that reception of Type-II
channels in the same group all begins and ends at the same time as shown in Figure 18.6.
Consequently, individual Type-II channels in the same group no longer need to be multicast
over a separate network channel, but can be transmitted over a single shared channel.
Let n 1 be the total number of Type-I channels and n 2 , j be the number of Type-II channels in
group j ( j = 0 , 1 ,... ) respectively. Then the bandwidth allocation of each channel in group j ,