Multilayer network survivability

1
Multilayer network survivability fault tolerant
VPN design

• Yu Liuyuliu_at_tele.pitt.edu
• OPNET Technologies, Inc.
• Joint work with Dr. David Tipper at
• University of Pittsburgh

2
Multi-layer networks

• Current network architect is multi-layer
• Technology, management
• IP/MPLS, ATM, FR, SONET/SDH, WDM, Physical
• Problems for survivable network design
• Topology layout
• Maintain upper layer connectivity for lower layer
  failures
• Capacity Allocation and Flow Assignment
• Reduce Cost/performance ratio
• Spare capacity allocation (include backup path
  assignment)

3
1. How to layout upper links

• Decision Interlayer information matrix H
• What is the best H to
• Avoid any lower layer failures
• Minimize total capacity used at lower layer

Columns are indexed by lower layer links
Rows are indexed by upper layer links
4
A matrix based layout model

• Based on Modianos work but extended to include
• matrix, and prevent arbitrary lower layer
  failures
• min eTH e
• s.t. H Bl T BT 0
• C Hf lt C e
• Hf H Fl T
• C is the cut-set link incidence matrix of upper
  layer
• Bl and B are node-link incidence matrices for
  lower and upper layers
• Fl is the failure-link incidence matrix at lower
  layer

Minimize the total trunks used
Flow conservation
Mengers theorem at upper layer
Failure propagation matrix
5
Cut-set link matrix C

• Each cutset is a row in C
• Each link is a column in C
• Total number of cutset 2N-1
• Problem not scalable , NP
• Cplex to solve small cases

L Link to be included in the failure
C
CutSets
6
2. Spare Capacity Allocation

• Given working paths and network topology
• Provision spare capacity and find backup routes
• Goal minimize total spare capacity or cost
• Assumption
• Use path restoration with disjoint backup routes
• Guarantee 100 protection for all single link
  failures
• Previous Work
• Integer Programming -- NP Hard
• Heuristics -- Quickly find feasible

7
Find spare capacity s
i
1
2
3
4
5
6
7
Link
From working and backup paths, G QT P
Backup path link
incident matrix
G
QT
s
1
2
0
2
1
1
1
0
1
0
0
1
1
0
1
1
0
0
0
2
2
2
0
2
1
1
0
0
1
1
0
0
0
1
1
0
0
0
3
1
0
1
0
0
0
1
1
0
0
1
0
0
0
0
0
1
0
4
1
1
1
1
0
0
1
0
1
0
0
1
1
0
0
0
1
0
5
2
1
1
1
0
0
0
2
0
1
1
0
0
0
0
0
0
1
From G, smaxG
6
1
0
0
1
0
1
0
1
0
0
0
0
1
0
0
1
0
1
7
2
1
0
2
0
2
0
0
0
1
0
0
0
1
0
1
0
0
11
Flows
1
2
5
6
7
8
9
10
3
4
src
dst
a
b
1
0
0
0
0
0
0
1
P
a
c
1
0
1
0
0
0
0
2
a
d
0
1
0
0
0
0
1
3
0
8
IP Model
Total spare capacity

• min S eT s
• s.t. s ? G
• G QT M P
• P Q ? 1
• Q BT D (mod 2)
• Q is a binary matrix
• Decision variable Q, G, s
• Given P working path link incidence matrix
• B and D node-link flow-node incidence matrices

Enough spare capacity on each link
Calculation of spare provision matrix
Link-disjointed backup paths
Flow conservation of backup
Integer programming
9
Arbitrary failure

• Consider each failure will affect all its
  adjacent links, use U, T to replace P
• U P FT , binary multiply, capture logical
  relations
• T U F

L Links
G
N Node failures
L Links
G
Q
U
L Link failures
Q

P
P
R Flows
R Flows
FT
SCA structure for single link failures
SCA structure for arbitrary failures
10
Consider arbitrary failures
Total spare capacity

• min S eT s
• s.t. s ? G
• G QT M U
• T Q ? 1
• Q BT D (mod 2)
• Q is a binary matrix
• U P FT
• T U F
• Decision variable Q, G, s

Enough spare capacity on each link
Calculation of spare provision matrix
Failure-disjointed backup paths
Flow conservation of backup
Integer programming
Path failure incident matrix
Path tabu-link incident matrix
Binary matrix multiplication
11
3 Multilayer SCA

• Restoration in lower layer only
• Each upper layer link is a lower layer demand
• Pl H
• Easier, treated as a single layer SCA problem
• Restoration in upper layer only
• Each lower layer link is an upper layer failure
• F HfT, interlayer failure propagation
  considered
• To be discussed in the following models
• Restoration cooperation between two layers
• Both layers have flows, sharing capacity across
  layers

12
Model 1 unaware layout

• Interlayer information H is used to convert spare
  capacity only
• min S1 eT sl eT HT s fT maxG1
• Other constraints are same as single layer case
• Problem
• Can not deal with failure propagation, can not
  provide enough connectivity and capacity upon
  failures

13
Model 2 use layout as failure

• Interlayer information H is used to consider
  failure propagation
• F Hf T Fl HT
• U P FT P H Fl T
• G2 QTMU
• min S2 eT sl eT HT s
• Advantage
• Guaranteed bandwidth under lower layer failures
• S1 ? S2

14
Model 3 use layout as failure and spare capacity
sharing

• Interlayer information H is used to consider
  failure propagation, and share spare capacity at
  the granularity of lower layer links
• min S3 eT sl eT max (HT G2)
• SSR modification
• vr H vrl
• Pros and cons
• S3 ? S2
• Require to control of link reservation at lower
  layer

share spare capacity on a-e
15
Networks

• Network 1 Network 2
• 5 nodes 18 nodes

16
VPN design examples

• H is obtained from the layout model earlier
• Branch and bound (BB) and SSR are used
• SSR find near optimal solution and fast

17
VPN design example 2

• Use Model 2 to compare
• Path restoration schemes
• Failure Independent (FID)
• Failure Dependent (FD)
• FD with Stub Release (FDStub)
• Survivable level
• Link failure
• Node failure

18
Discussion and Questions

• Applications
• IP/MPLS over Optical
• Leased VPN
• Spare sharing between two layers
• Model is derived in the paper
• Data not discussed in this talk
• Single Layer SCA tomorrow TB01.1
• http//www2.sis.pitt.edu/yliu/