Information Technology Reference
In-Depth Information
Server
S
0
S
1
S
2
S
3
S
4
Stripe
v
0
v
1
v
2
v
3
v
4
v
5
v
6
v
7
v
8
v
9
Stripe unit
v
10
v
11
v
12
v
13
v
14
v
15
v
16
v
17
v
18
v
19
v
20
v
21
v
22
v
23
v
24
v
i
is stripe unit
i
, containing frames
ki
to
k
(
i
+1) −1
Figure 9.4
Striping a video stream over five servers using time striping
Existing studies using the time striping approach include those works by Biersack
et al.
[8, 9], and Buddhikot and Parulkar [11]. The study by Bernhardt and Biersack [9] adopted
time striping with a granularity of one frame and also one segment of a frame -
sub-frame
striping
. For sub-frame striping, a frame is sub-divided into
k
equal-size units and then dis-
tributed across the servers in a round-robin fashion. The key advantage of sub-frame striping
is that load balance is guaranteed for both CBR and VBR video streams as each frame is
striped equally across all servers. Conversely, the study by Buddhikot and Parulkar [11] used
a stripe unit of
k
(
k
1) frames. They suggested solving the load balance problem by group-
ing more frames into a stripe unit to obtain a more uniform stripe unit size (see Section
9.3.4).
≥
9.3.2 Space Striping
Time striping divides a video stream into fixed-length (in time) stripe units. Another approach
is to divide a video stream into fixed-size (in bytes) stripe units -
space striping
. Space striping
simplifies storage and buffer management in the servers because all stripe units are of the same
size. Moreover, the amount of data sent by each server in a service round is also the same.
Unlike time striping, we do not need to know the frame structure in order to perform striping,
thus decoupling the striping algorithm from the encoding format.
This space striping approach has been employed in many studies [10, 12-18]. Depending
on the system design, the stripe unit size can range from tens of kilobytes to hundreds of
kilobytes. In most of the studies, a stripe unit is assumed to play back in a constant length of
time. However, in most video compression algorithms such as MPEG, a fixed-size stripe unit
will likely contain a variable number of frames and/or partial frames. Moreover, if the video
is compressed using constant-quality compression algorithms (cf. Section 2.4), then the video
bit-rate will become variable as well. Consequently, the decoding time for a stripe unit at the
Search WWH ::
Custom Search