Seamless Detection of Link and Node Failures for Local Protection in MPLS

1
Seamless Detection of Link and Node Failures for
Local Protection in MPLS

• Zartash Afzal Uzmi
• Computer Science and Engineering
• Lahore University of Management Sciences (LUMS)
• Visiting Professor Chonbuk National University

2
Outline

• Background
• Forwarding and Routing in IP and MPLS Networks
• Network Service Requirements
• Protection Routing in MPLS
• Terminology Types of Backup Paths
• Backup Bandwidth Sharing
• Activation sets
• Failures and Backup Path Activation
• Distinguishable Failure Events Ideal Case
• Actual Failures
• Control Plane Mechanism
• Outline of Proof

3
Outline

• Background
• Forwarding and Routing in IP and MPLS Networks
• Network Service Requirements
• Protection Routing in MPLS
• Terminology Types of Backup Paths
• Backup Bandwidth Sharing
• Activation sets
• Failures and Backup Path Activation
• Distinguishable Failure Events Ideal Case
• Actual Failures
• Control Plane Mechanism
• Outline of Proof

4
Forwarding and Routing

• Forwarding
• Passing a packet to the next hop router
• Routing
• Computing the best path to the destination
• IP routing includes routing and forwarding
• Each router makes the routing decision
• Each router makes the forwarding decision
• IP routing is hop-by-hop
• MPLS routing
• Only one router (source) makes the routing
  decision
• Intermediate routers make the forwarding decision
• An MPLS path or virtual circuit from source to
  destination is created and is called an LSP
  (label switched path)

5
Network Service Requirements

• Bandwidth Guaranteed Primary Paths
• MPLS can establish bandwidth-guaranteed paths
• Bandwidth Guaranteed Backup Paths
• BW remains provisioned in case of network failure
• Two options for recovery from network failure
• Compute backup paths AFTER failures occur
• Compute and install PRESET backup paths
• Minimal Recovery Latency
• Recovery latency is the time that elapses
  between
• the occurrence of a failure, and
• the diversion of network traffic on a new path

Preset backup paths needed for minimal latency
6
Outline

• Background
• Forwarding and Routing in IP and MPLS Networks
• Network Service Requirements
• Protection Routing in MPLS
• Terminology Types of Backup Paths
• Backup Bandwidth Sharing
• Activation sets
• Failures and Backup Path Activation
• Distinguishable Failure Events Ideal Case
• Actual Failures
• Control Plane Mechanism
• Outline of Proof

7
Protection in MPLSPreset Backup Paths
Local Protection
Path Protection
S
1
2
3
D
This type of path Protection takes 100s of ms.
We need Local Protection to quickly switch onto
backup paths!
Primary Path
Backup Path
8
nhop and nnhop paths
LOCAL PROTECTION (showing one LSP only)
All links and all nodes are protected!
nnhop
A
B
D
C
E
nhop
PLR Point of Local Repair
? nhop protects link only, e.g., (D,E) ? nnhop
protects link (C,D) and node (D)
9
Opportunity cost of backup paths

• Protection requires that backup paths are setup
  in advance
• Upon failure, traffic is promptly switched onto
  preset backup paths
• Bandwidth must be reserved for all backup paths
• This results in a reduction in the number of
  Primary LSPs that can otherwise be placed on the
  network
• Can we reduce the amount of backup bandwidth
  but still provide guaranteed backups?
• YES Try to share the bandwidth along backup paths

10
BW Sharing in backup Paths

• Example

LSP1
BW X
Sharing is possible IF Links (A,B) and (C,D) do
not simultaneously fail!
A
B
X
X
max(X, Y)
X
E
G
F
XY
Y
Y
C
D
BW Y
LSP2
11
Activation Sets
Can backup paths always share the bandwidth?
A
A
E
E
B
B
C
C
D
D
Activation set for node B
Activation set for link (A,B)
backup paths in the same activation set MUST not
share the bandwidth!
12
Outline

• Background
• Forwarding and Routing in IP and MPLS Networks
• Network Service Requirements
• Protection Routing in MPLS
• Terminology Types of Backup Paths
• Backup Bandwidth Sharing
• Activation sets
• Failures and Backup Path Activation
• Distinguishable Failure Events Ideal Case
• Actual Failures
• Control Plane Mechanism
• Outline of Proof

13
Distinguishable Failure Events
Point of local repair (PLR) somehow knows the
type of failure!
Focus on link (I,J) and Node J and recall
? nhop protects link only i.e., (I,J) ? nnhop
protects link (I,J) and node J
nnhop p1
A
J
I
K
nhop p2
PLR Point of Local Repair
L
p3
If node I finds that link (I,J) has failed p1
and p2 are activated If node I finds that node J
has failed ONLY p1 is activated
p2 may share bandwidth with other nnhops that
protect node j
14
Actual Failures

• Consider the failure of link (I,J)
• Both p1 and p2 need to be activated, anyways!
• Knowing that this is a link failure will not save
  anything
• Consider the failure of node J
• Only p1 needs to be activated (if failure type is
  known!)
• What if node I doesnt know the type of failure?
• Two options
• Wait to discover if it was a link or node
  failure
• High recovery latency (BAD!)
• Activate both p1 and p2 instantaneously
• Now p2 will not be able to share with p3 (BAD!)

15
Control Plane Mechanism

• Routing strategy
• Do not oversubscribe
• Use sharing as if adjacent nodes can distinguish
  the node failures from the link failures
• That is, provide sharing between p2 and p3
• In reality
• PLRs will not be able to disambiguate link/node
  failures
• Activate p1 and p2 (assuming link fail situation
  worst case!)
• If link had failed
• p1 and p2 really needed to be activated we are
  okay!
• If node had failed
• p2 (nhop) has been activated by mistake
• You may notice reservation violation at some
  nodes (where the backup paths p2 and p3 were
  sharing)
• Abort all nhop paths that are violating the
  reservations

16
Outline of Proof

• Define
• Guv Bandwidth reserved on link (u, v) for all
  backup LSPs
• Iuv Actual backup bandwidth that falls on link
  (u, v), after the occurrence of a failure
• A reservation violation happens if Iuv gt Guv
• No oversubscription sharing between p2 and p3
• Guv max(bw(p1)bw(p2), bw(p1)bw(p3)) worst
  case
• When failure occurs, activate p1 and p2
• If it was link (I, J) that had failed, we are
  okay
• If it was node J that had failed, p3 also gets
  activated
• Worst case Iuv would have been bw(p1)bw(p2)bw(p3
  )
• Our control plane mechanism ensures Iuv
  bw(p1)bw(p3)
• This implies that Guv Iuv in the worst case

17

• Questions Answers