Algorithms for computing Maximally Redundant Trees for IP

Algorithms for computing Maximally Redundant
Trees for IP/LDP Fast-Reroute
draft-enyedi-rtgwg-mrt-frr-algorithm-01
Alia Atlas (Juniper Networks)
Gabor Enyedi, Andras Csaszar (Ericsson)
IETF 83, Paris, France
1
Agenda
•
•
•
•
Briefly about the algorithm
Problem
Avoid using a node
Non-2-connected networks
2
Agenda
•
•
•
•
Briefly about the algorithm
Problem
Avoid using a node
Non-2-connected networks
3
ADAG and partial order
E
G
D
Root
C
S
B
A
H
4
ADAG and partial order
• Almost DAG (ADAG)
• A<<B if there is a path from A to B
• Root is both the shortest and the greatest
E
G
D
Root
C
S
B
A
H
5
ADAG and partial order
• S<<E
– Blue path: increasing [S, E]
– Red path: decreasing [Root, S] and [E, Root]
E
G
D
Root
C
S
B
A
H
6
ADAG and partial order
• S>>A
– Blue path: increasing [S,Root] and [Root, A]
– Red path: decreasing [A, S]
E
G
D
Root
C
S
B
A
H
7
ADAG and partial order
• S and C are not ordered
– Blue path: [S, E] and [C, E]
– Red path: [A, S] and [A, C]
E
G
D
Root
C
S
B
A
H
8
Agenda
•
•
•
•
Briefly about the algorithm
Problem
Avoid using a node
Non-2-connected networks
9
Three trees
• We have tree trees
– SPT
– Two MRTs
• There is no connection between SPT and MRTs
• Impossible to find a redundant pair for SPT
• Example:
Shortest path
1
C
D
10
1
S
Dest
1
10
No redundant
pair for that!
1
B
A
1
10
Agenda
•
•
•
•
Briefly about the algorithm
Problem
Avoid using a node
Non-2-connected networks
11
Total order
• Partial order can compare any X only with S
– We need to compare any two nodes
• Make a total order as well
– If A<<B, let A<B
– If A and B are not ordered select either A<B or B<A
– This can be done with a topological oder after
converting the ADAG into a DAG
• Results:
• If A<B, either A<<B or A and B are not ordered
12
A possible total order
• Numbers are written next to nodes
8
7
E
G
D
Root
6
C
5
0
S
B
1
A
4
3
H
2
13
Possible cases
• If dst>>src, failed node F
8
7
E
G
D
Root
If S<<F<E, it may
be on the BLUE
path
6
C
5
0
S
B
1
A
4
3
Otherwise: it may
be on the RED path
H 2
14
Possible cases
• If dst<<src, failed node F
Otherwise: it may
be on the BLUE
path
7
8
E
G
D
Root
6
C
5
0
S
4
3
B it may be
If A<F<<S,
on the RED path
1
A
H
2
15
Possible cases
• If dst and src are not ordered
If F>>src, it may
be on the first part
of the RED path
– There are four sub-paths
If F and src are not
8
ordered, and F>dest, E
it may be on the
second part of the
RED path
Root
5
7
G
6
D
C
S
0
F and src are not
ordered, and F<dest, it
may be on the second
part of BLUE path 1 A
4
3
B
H
2
If F<<src, it may
be on the first part16
of the BLUE path
Agenda
•
•
•
•
Briefly about the algorithm
Problem
Avoid using a node
Non-2-connected networks
17
Non-2-connected problem
• In this case we don’t have a single order
– Neither a partial order
– Nor a total order
• Convert the GADAG into an ADAG!
C
A
Non-root block
B
X
Local root block
D
18
Non-2-connected problem
• In this case we don’t have a single order
– Neither a partial order
– Nor a total order
• Convert the GADAG into an ADAG!
C
A
X1
Non-root block
Local root block
X2
B
D
19
Thank you!
20