A Staggered-Push Parallel Server Architecture - Scalable Continuous Media Streaming Systems - page 226

Again noting that max {| d i − d j ||∀ i , j }= τ , the worst case is

f + −

f − +

T L + τ

Z > 1 + (12.32)

T F

which is the number of empty buffers needed to avoid client buffer overflow.

12.5.3 System Response Time
Another key performance metric of a VoD service is system response time, defined as the time
from initiating a new request to the time video playback starts. Ignoring system administration
and network delay issues, system response time consists of two components: scheduling delay
and prefill delay. Scheduling delay is the delay incurred at the admission scheduler plus the
delay incurred at the server scheduler, as derived in Section 12.3. For prefill delay, we note that
the client prefills the first Y video blocks before starting video playback. Hence, the average
prefill delay can be obtained from

YQ
R V

D P = (12.33)

and the system response time is simply the sum D S + D P .

12.6 Performance Evaluation
In this section, we evaluate the performance of the staggered-push architecture using numerical
results. All results are computed using the derivations in Sections 12.3 to 12.5 with the system
parameters listed in Table 12.1.

12.6.1 Design Example
To illustrate performance and resource requirements of the architecture, we first consider a
design example in this section. We assume that there are 8 servers in the system, with a
client-server ratio of 50 (i.e., up to 400 concurrent streams). Using the parameters in Table
12.1, the server buffer requirement is calculated to be 6.25MB. Compared to the amount of
memory in today's PC, this buffer requirement is relatively small. Moreover, as conventional
PCs can be expanded to 1GB or more memory, in theory a client-server ratio of over 8,000

Table 12.1 System parameters used in performance evaluation

System Parameters Symbol Value

Video block size Q 65,536 Bytes

Video data rate R V 150 KB/s

Maximum early in decoding time T E − 130 ms

Maximum late in decoding time T L 160 ms

Client-Server ratio 10

f − , f +

Transmission time deviation 0ms

Server clock jitter τ 100 ms