Soft Scheduling - Scalable Continuous Media Streaming Systems - page 78

160

140

120

100

80

60

40

20

0

0 0.5 1 1.5 2 2.5 3 3.5 4

Scheduling Delay (seconds)

Hard-Scheduling, w/o EAS
Soft-Scheduling, w/o EAS
Dual-Round Scheduling, w/o EAS
Hard-Scheduling, w/EAS
Soft-Scheduling, w/EAS
Dual-Round Scheduling, w/EAS

Figure 4.9 Usable disk capacity versus scheduling-delay constraint ( R = 150KB, ε = 10 − 6 )

constraints. This is explained by the fact that it is more likely to have sufficient slack time in
a round to absorb overflow when the overflow probability constraint is small. Given today's
low memory cost, DRS is an attractive option for achieving better capacity at the expense of
modest increase in buffer requirement.

4.6.4 Early-Admission Scheduling
To study the capacity gains from Early-Admission Scheduling, we compute the media block
size according to equation (4.22), and round it down to multiples of 64KB. The usable disk ca-
pacities for various combinations are shown in Figure 4.9. The horizontal axis is the admission
delay constraint used for computing the block size.
We observe that in all cases EAS can substantially increase the capacity, including both hard-
scheduling and soft-scheduling cases. For example, with an admission delay constraint of one
second, the usable disk capacity increases from 35 to 81 (131% increase) for hard scheduling
and increases from 47 to 107 (128% increase) for soft scheduling. The improvement in DRS
is similar. These dramatic increases in usable disk capacity are explained by the fact that the
admission delay is reduced by two-thirds under EAS. Therefore, substantially larger block size
can be used to improve disk efficiency.