Database Reference
In-Depth Information
Table 6.1 Execution time and accuracy rate of minimum replication
algorithm
One replica
Two replicas
Average execution time (ms)/(number of checking interval values)
Reliability
99%
99.9%
99.99%
99.999%
ORC
0.69
15.34
15.62
16.20
OPC
0.69
0.69
0.69
0.69
ORV
0.72 (1)
16.26 (1)
16.30 (1)
155.82 (10)
OPV
0.72 (1)
4 (1)
7.81 (2)
41.52 (11)
Accuracy rate
OPC
NA
89.52%
99.00%
99.90%
OPV
NA
89.61%
99.00%
99.90%
or not optimized, with the increment of the data reliability requirement, the execution
time of the variable disk failure versions of the algorithm increases fast. The reason
for this is because with a higher data reliability requirement, the checking interval of
the data file becomes shorter, so that during the expected storage duration of the data
file, more checking interval values need to be calculated. In the accuracy rate part of
Table 6.1 , due to storage with a single replica, the accuracy rate for data reliability of
99% is not applicable. Despite of that, the lowest accuracy rate is 89.52% when the
data reliability requirement is 99.9%. The accuracy rates increase with the increment
of the data reliability requirement, which is consi s tent with the trend of deviation be-
tween the tangent line and the original curve of
e t as shown in Figure 6.5 . According
to the existing results shown in Table 6.1 , the accuracy rates of the optimized versions
of the algorithm reach 99.9%. In fact, this value can be even larger when the data reli-
ability requirement becomes higher.
The results in Table 6.1 show that, depending on the data reliability assurance pro-
vided, the minimum replication algorithm is able to calculate the checking interval
values of data files between a few milliseconds to hundreds of milliseconds. However,
the reliability assurance is not restricted to that shown in the table, which can be even
higher (e.g. the data reliability assurance of 99.9999% is the same as that of the con-
ventional three-replica strategy) and easily changed according to the data reliability
requirement. To provide higher data reliability assurance, more time could be taken
to conduct the minimum replication algorithm, and more checking interval values
need to be calculated. The execution time of the optimized versions of the algorithm
is generally much shorter than that of the original versions, but the accuracy rate is
somewhat lower when the reliability requirement is lower and increases while the data
reliability requirement increases.
λ
6.5.1.2 Metadata scanning and proactive replica checking
An experimental PRCR is implemented based on the AWS, including Amazon Simple
Storage Service (Amazon S3), Amazon Elastic Compute Cloud (Amazon EC2), and
 
Search WWH ::




Custom Search