Database Reference
In-Depth Information
6.5.1 Performance of PRCR
Instead of being used to describe the speed of data access, the term “performance”
of PRCR refers to the running overhead as well as the maximum capacity of PRCR,
which are the key indicators that show the ability of cost-effectively managing the
Cloud data files. To evaluate the performance of PRCR, despite the time for creat-
ing replicas, we find that the minimum replication algorithm, the metadata scanning
procedure, and the proactive replica checking procedure are the major procedures that
most affect the performance of PRCR. Therefore, investigations of these three proce-
dures are conducted.
6.5.1.1 The minimum replication algorithm
We evaluate the minimum replication algorithm because of two reasons:
•
First, this evaluation is considered as the evaluation for the minimum replication approach
described in
Chapter 5
.
•
Second, as part of the user interface component in PRCR, this algorithm is of great signifi-
cance for conducting the first and second steps of the PRCR working process as depicted in
Figure 6.2
.
The evaluation is conducted on aspects of execution time and result accuracy. In
addition, as presented in
Chapter 5
, optimization approaches have been conducted for
simplifying the calculation process. In order to fully investigate the minimum replica-
tion algorithm and the effect of our optimization, the evaluation is carried out as fol-
lows: four versions of the algorithm are implemented, which are the original constant
disk failure rate version (version ORC), the optimized constant disk failure rate ver-
sion (version OPC), the original variable disk failure version (version ORV), and the
optimized variable disk failure version (version OPV). The original versions (ORC
and ORV) of the algorithm calculate the checking interval by solving equation
(6.3)
:
−
λ
t
−
λ
t
(1)
1 1
=
−− −
e
)(1
e
)
1
2
RA
(6.3)
k
t
which is the revised version of equation
(5.7)
. This equation calculates the checking
interval in the unoptimized form. Meanwhile, while the optimized versions (OPC and
OPV) calculate the checking interval by solving equation
(6.1)
. The constant disk
failure rate versions (ORC and OPC) of the algorithm are for storage with a constant
disk failure rate, while the variable disk failure rate versions (ORV and OPV) are for
storage with a variable disk failure rate. The evaluation of the constant disk failure
rate versions of the algorithm corresponds to the discussion in
Section 6.4
about the
algorithm working in a constant failure rate environment.
In equation
(6.1)
we
use the tangent line of
−
λ
t
at point (0
,
1) as a
s
ubstitution
e
−
λ
t
.
Th
e function of the tangent line is
for the original curv
e
e
ft t
()1
λ
=−
λ
, which is
()
mentioned in
Se
ction 5.1
, while a =
−
1 and b = 1.
Figure 6.5
sh
ows b
oth
the original curve of
a special case of
λ=+
ftat b
−
λ
t
and the substitution curve of tan-
e
gent line
ft t
()1
λ
=−
λ
. In the f
ig
ure, the substitution curve is located at the lower
side of the original curve of
Search WWH ::
Custom Search