Increasing disaster resiliency by having a pod backed up to other peer pods in a site or beyond
View Patent ↗Increasing disaster resiliency in one aspect may comprise running an optimization algorithm that simultaneously solves for at least a first objective to increase a spread of a backup of virtual machines from a given site onto other sites in proportion to an amount of available space for backup at each site, a second objective to increase a number of backups at one or more of the other sites with low probability of system crash while reducing backups at one or more of the other sites with higher probability of system crash, and a third objective to minimize a violation of recovery time objectives of the virtual machines during recovery. One or more backup sites and one or more recovery sites in an event the given site crashes may be determined based on a solution of the optimization algorithm.
1. A method of increasing disaster resiliency in computer systems, comprising:
executing, by a processor, an optimization algorithm that solves simultaneously for at least a first objective to increase a spread of a backup of virtual machines from a given site onto other sites in proportion to an amount of available space for backup at each site, a second objective to increase a number of backups at one or more of the other sites with low probability of system crash while reducing backups at one or more of the other sites with higher probability of system crash, and a third objective to minimize a violation of recovery time objectives of the virtual machines during recovery; and
determining one or more backup sites and one or more recovery sites, in an event the given site crashes, based on a solution of the optimization algorithm.
2. The method of claim 1 , wherein the given site and the other sites comprise points of delivery that comprise hosting environments where the virtual machines belonging to one or more applications run.
3. The method of claim 1 , wherein the optimization algorithm is solved subject to a processing capacity constraint associated with at least the other sites, storage capacity constraint associated with at least the other sites, and data privacy constraints associated with at least the virtual machines to be backed up.
4. The method of claim 1 , further comprising constructing a schedule of backups for the virtual machines on one or more of the other sites based on a solution of the optimization algorithm.
5. The method of claim 1 , further comprising constructing a schedule of recovery for the virtual machines on one or more of the other sites based on a solution of the optimization algorithm.
6. The method of claim 1 , wherein the first objective is formulated as:
minimize
∑
i
=
1
…
N
∑
(
r
,
s
)
,
r
≠
s
≠
i
,
r
,
s
∈
{
1
…
N
}
(
A
s
∑
j
=
1
…
n
i
x
rij
-
A
r
∑
j
=
1
…
n
i
x
sij
A
r
+
A
s
)
2
,
wherein
sites comprising the given site and the other sites are indexed by iε{1, 2, . . . , N},
n i represents a number of VM instances hosted at site i,
s ij represents a rate of backup flow for VM j hosted at site i,
A i represents available space at site i for keeping backups from other sites,
X kij =1 if VM j hosted at site i is backed up at site k, otherwise 0.
7. The method of claim 1 , wherein the first objective is formulated as:
minimize
minimize
∑
i
=
1
…
N
∑
(
r
,
s
)
,
r
≠
s
≠
i
,
r
,
s
∈
{
1
…
N
}
(
∑
j
=
1
…
n
i
x
rij
-
∑
j
=
1
…
n
i
x
sij
)
,
wherein
sites comprising the given site and the other sites are indexed by iε{1, 2, . . . , N},
n i represents a number of VM instances hosted at site i,
s ij represents a rate of backup flow for VM j hosted at site i,
x kij =1 if VM j hosted at site i is backed up at site k, otherwise 0.
8. The method of claim 1 , wherein the second objective is formulated as:
∑
i
=
1
…
N
∑
(
r
,
s
)
,
r
≠
s
≠
i
,
r
,
s
∈
{
1
…
N
}
(
f
(
p
r
)
∑
j
=
1
…
n
i
x
rij
-
f
(
p
s
)
∑
j
=
1
…
n
i
x
sij
f
(
p
r
)
+
f
(
p
s
)
)
2
,
wherein
function f represents a crash probability function,
sites comprising the given site and the other sites are indexed by iε{1, 2, . . . , N},
n i represents a number of VM instances hosted at site i,
s ij represents a rate of backup flow for VM j hosted at site i,
x kij =1 if VM j hosted at site i is backed up at site k, otherwise 0.
9. The method of claim 1 , wherein the third objective is formulated as:
max
i
=
1
…
N
max
r
=
1
…
N
,
r
≠
i
∑
c
=
1
…
M
p
i
ψ
c
(
v
ric
)
,
wherein
ψ c (.) is a normalized penalty function for recover time objective violation of application c,
sites comprising the given site and the other sites are indexed by iε{1, 2, . . . , N},
p i represents probability that site i crashes,
n i represents a number of VM instances hosted at site i,
s ij represents a rate of backup flow for VM j hosted at site i,
x kij =1 if VM j hosted at site i is backed up at site k, otherwise 0.
10. The method of claim 1 , wherein the first objective further comprises a first weight value associated with the first objective, the second objective further comprises a second weight value associated with the second objective, or the third objective further comprises a third weight value associated with the third objective, or combinations thereof.
11. The method of claim 1 , wherein the optimization algorithm comprises:
minimize
w
1
∑
i
=
1
…
N
∑
(
r
,
s
)
,
r
≠
s
≠
i
,
r
,
s
∈
{
1
…
N
}
(
A
s
∑
j
=
1
…
n
i
x
rij
-
A
r
∑
j
=
1
…
n
i
x
sij
A
r
+
A
s
)
2
+
w
2
∑
i
=
1
…
N
∑
(
r
,
s
)
,
r
≠
s
≠
i
,
r
,
s
∈
{
1
…
N
}
(
f
(
p
r
)
∑
j
=
1
…
n
i
x
rij
-
f
(
p
s
)
∑
j
=
1
…
n
i
x
sij
f
(
p
r
)
+
f
(
p
s
)
)
2
+
w
3
max
i
=
1
…
N
max
r
=
1
…
N
,
r
≠
i
∑
c
=
1
…
M
p
i
ψ
c
(
v
ric
)
,
wherein
∑
i
=
1
…
N
∑
(
r
,
s
)
,
r
≠
s
≠
i
,
r
,
s
∈
{
1
…
N
}
(
A
s
∑
j
=
1
…
n
i
x
rij
-
A
r
∑
j
=
1
…
n
i
x
sij
A
r
+
A
s
)
2
represents the first objective and w1 represents a configurable weight associated with the first objective,
∑
i
=
1
…
N
∑
(
r
,
s
)
,
r
≠
s
≠
i
,
r
,
s
∈
{
1
…
N
}
(
f
(
p
r
)
∑
j
=
1
…
n
i
x
rij
-
f
(
p
s
)
∑
j
=
1
…
n
i
x
sij
f
(
p
r
)
+
f
(
p
s
)
)
2
represents the second objective and w2 represents a configurable weight associated with the second objective, and
max
i
=
1
…
N
max
r
=
1
…
N
,
r
≠
i
∑
c
=
1
…
M
p
i
ψ
c
(
v
ric
)
represents the third objective and w3 represents a configurable weight associated with the third objective.
12. A non-transitory computer readable storage device storing a program of instructions executable by a machine to perform a method of increasing disaster resiliency in computer systems, comprising:
executing, by a processor, an optimization algorithm that solves simultaneously for at least a first objective to increase a spread of a backup of virtual machines from a given site onto other sites in proportion to an amount of available space for backup at each site, a second objective to increase a number of backups at one or more of the other sites with low probability of system crash while reducing backups at one or more of the other sites with higher probability of system crash, and a third objective to minimize a violation of recovery time objectives of the virtual machines during recovery; and
determining one or more backup sites and one or more recovery sites in an event the given site crashes based on a solution of the optimization algorithm.
13. The non-transitory computer readable storage device of claim 12 , wherein the given site and the other sites comprise points of delivery that comprise hosting environments where the virtual machines belonging to one or more applications run.
14. The non-transitory computer readable storage device of claim 12 , wherein the optimization algorithm is solved subject to a processing capacity constraint associated with at least the other sites, storage capacity constraint associated with at least the other sites, and data privacy constraints associated with at least the virtual machines to be backed up.
15. The non-transitory computer readable storage device of claim 12 , further comprising constructing a schedule of backups for the virtual machines on one or more of the other sites based on a solution of the optimization algorithm.
16. The non-transitory computer readable storage device of claim 12 , further comprising constructing a schedule of recovery for the virtual machines on one or more of the other sites based on a solution of the optimization algorithm.
17. A system for increasing disaster resiliency in computer systems, comprising:
an optimization model that solves simultaneously for at least a first objective to increase a spread of a backup of virtual machines from a given site onto other sites in proportion to an amount of available space for backup at each site, a second objective to increase a number of backups at one or more of the other sites with low probability of system crash while reducing backups at one or more of the other sites with higher probability of system crash, and a third objective to minimize a violation of recovery time objectives of the virtual machines during recover; and
a hardware processor operable to execute the optimization model to determine one or more backup sites and one or more recovery sites in an event the given site crashes based on a solution of the optimization model.
18. The system of claim 17 , further comprising:
a memory storing the optimization model, wherein the optimization model solves:
minimize
w
1
∑
i
=
1
…
N
∑
(
r
,
s
)
,
r
≠
s
≠
i
,
r
,
s
∈
{
1
…
N
}
(
A
s
∑
j
=
1
…
n
i
x
rij
-
A
r
∑
j
=
1
…
n
i
x
sij
A
r
+
A
s
)
2
+
w
2
∑
i
=
1
…
N
∑
(
r
,
s
)
,
r
≠
s
≠
i
,
r
,
s
∈
{
1
…
N
}
(
f
(
p
r
)
∑
j
=
1
…
n
i
x
rij
-
f
(
p
s
)
∑
j
=
1
…
n
i
x
sij
f
(
p
r
)
+
f
(
p
s
)
)
2
+
w
3
max
i
=
1
…
N
max
r
=
1
…
N
,
r
≠
i
∑
c
=
1
…
M
p
i
ψ
c
(
v
ric
)
,
wherein
∑
i
=
1
…
N
∑
(
r
,
s
)
,
r
≠
s
≠
i
,
r
,
s
∈
{
1
…
N
}
(
A
s
∑
j
=
1
…
n
i
x
rij
-
A
r
∑
j
=
1
…
n
i
x
sij
A
r
+
A
s
)
2
represents the first objective and w1 represents a configurable weight associated with the first objective,
∑
i
=
1
…
N
∑
(
r
,
s
)
,
r
≠
s
≠
i
,
r
,
s
∈
{
1
…
N
}
(
f
(
p
r
)
∑
j
=
1
…
n
i
x
rij
-
f
(
p
s
)
∑
j
=
1
…
n
i
x
sij
f
(
p
r
)
+
f
(
p
s
)
)
2
represents the second objective and w2 represents a configurable weight associated with the second objective, and
max
i
=
1
…
N
max
r
=
1
…
N
,
r
≠
i
∑
c
=
1
…
M
p
i
ψ
c
(
v
ric
)
represents the third objective and w3 represents a configurable weight associated with the third objective.