Information Technology Reference
In-Depth Information
Figure 5.5 Track-based automatic data rebuild.
simply the time for one disk rotation, i.e., W 1 . After reading the corresponding tracks from
all ( N D
1) disks, the server can then reconstruct the lost track and write it to the spare disk.
Additionally, the rebuild process rebuilds tracks sequentially starting from one end of the
disk surface with all track retrievals performed back-to-back in one go. For example, let
y i ,
j , be the track numbers for the u data blocks to
be retrieved for playout in a round. Suppose that the next track to rebuild is track number x
and a total of
i
=
0
,
1
,...,
( u
1) and y i
y j for i
<
tracks are to be rebuilt. Then the order of retrievals will be y 0 , y 1 , ..., y i , x ,
y j ,..., y u 1 , where y i
v
tracks are retrieved in one go between
the retrievals of block i and j . Consequently, the seek time between track retrievals is reduced
to t seek (1). The rebuild process will retrieve as many tracks as possible in a round for rebuild
as long as retrieval performance for normal data blocks can still be guaranteed.
x
y j . In other words, all
v
5.5.2 Analysis of Rebuild Time
To model the rebuild process, let u ( u
K ) be the number of media blocks to retrieve for
playback and
v
be the number of tracks to retrieve for rebuild in a service round. Using the
Search WWH ::




Custom Search