Information Technology Reference
In-Depth Information
Replication with Checkpointing
The execution time for replication with checkpointing is as follows:
(
)
l
/(
NT TT
-
+
)
l
/(
NTT
+
)
(4.21)
e
e
-
1
ʈ
=
Á˜
T
RD C R
RI
C
Q
F
et rep ch
(
+
)
˯
T
l
/
N
I
R
The execution cost for replication with checkpointing is computed as
Q
co
=
Q
price
¥
Q
et(rep
+
ch)
¥
N
R
(4.22)
Retrying with Replication
The execution time for retrying with replication is
(
)
(l
/)
NT
(l
/)
NT
(4.23)
e
e
-
1
RD
RF
Q
=
et ret
(
+
rep
)
l
/
N
R
The execution cost for retrying with replication is computed as
Q
co
=
Q
price
¥
Q
et
(
ret
+
rep
)
¥
N
R
(4.24)
Policy-Making Model
The basic idea of adaptive model in DRIC is to set the optimized failure-
recovery method using the multiple attribute decision making
(MADM)
approach [73]. With the above dei nitions and the equations, the decision-
making matrix
M
is obtained from Equation 4.25 as
Q
Q
QQ
Ê
ˆ
Ê
ˆ
et ret
()
co ret
()
1,1
1,2
Á
˜
Á
˜
Q
Q
QQ
Á
et ch
()
co ch
()
˜
2,1
2,2
Á
˜
M
=
=
(4.25)
ij
,
Q
Q
QQ
Á
˜
Á
˜
et rep
(
+
ch
)
co rep
(
+
ch
)
3,1
3,2
Á
˜
Á
˜
Q
Q
QQ
Ë
¯
Ë
¯
et ret
(
+
rep
)
co ret
(
+
rep
)
4,1
4,2
As execution time and execution cost have different quantum, the matrix
M
need to be formalized with Equation 4.27.
Search WWH ::
Custom Search