Information Technology Reference
In-Depth Information
it is easy to see that
t f
t j
t f
+
T avg
(11.28)
Since the client will need to wait for the redundant unit before stripe j can be re-computed,
the delay incurred in receiving stripe j will be given by
t f
T avg
t f
T D
+
T D +
t j
+
t f
=
T D +
T avg
(11.29)
which, finally, is independent of the system scale. Using derivations similar to Section 11.5,
we can obtain the client buffer requirement from
3
τ +
τ +
Q
f +
f
f +
f +
T E +
T D
T L
B RSS
PRT
=
+
+
(11.30)
T avg
T avg
for the case where failure occurs after playback has begun and
3
τ +
τ +
Q
f +
f
f +
f +
T E
T L +
T D
B RSS
PRT =
+
+
(11.31)
T avg
T avg
for the case where failure occurs before playback begins.
To support immediate retransmission of redundant units, the redundant servers will need to
retain redundant units longer than in the original AGSS scheduler. In particular, the server will
need to keep retrieved redundant units (in blocks of N S units) for
T D
N S T avg
+
1
(11.32)
service rounds (instead of one round in AGSS). Hence, the buffer requirement for the redundant
servers will be given by
1
T D
N S T avg
1
G +
B server =
QN S
+
(11.33)
where
is the client-server ratio and G is the number of groups per service round.
11.7 Numerical Results
Based on the performance models derived in the previous sections, we compute and present
numerical results in this section to illustrate the system resource requirement under various
scenarios and study the sensitivity to key system parameters. Table 11.1 lists the values for
the system parameters used in the calculation. The parameters T E and T L are determined
empirically by collecting the video block consumption times of a hardware MPEG decoder.
Search WWH ::




Custom Search