A Diffserv-TE Implementation Model to dynamically change booking factors during failure events draft-newton-mpls-te-dynamic-overbooking-00.txt

1
A Diffserv-TE Implementation Model to
dynamically change booking factors
during failure eventsdraft-newton-mpls-
te-dynamic-overbooking-00.txt

• Jonathan Newton CableWireless
• Mustapha Aissaoui Alcatel-Lucent
• JP. Vasseur Cisco

2
Problem Statement

• Current MPLS-TE (and DSTE) implementations
  advertise a fixed maximum bandwidth for each
  resource.
• (Over)booking factors are fixed
• Network operates with a fixed behaviour.
• Requirement is to allow the behaviour of the
  network to change during the failure of a
  resource.
• Example engineer to 60 link utilisation under
  normal network operation but allow 80 link
  utilisation under a network failure event.

3
Proposed Solution

• Use DSTE to advertise different BCs
• One BC for normal operation
• One BC for failure operation.
• Use RDM, such that normal operation BC is
  nested within failure operation BC.
• Signal LSPs using corresponding CT.
• Identify failure mode at the HE when either
• LSP receives path-tear
• LSP receives relevant Path-Err (eg. Tunnel
  locally repaired)

4
Example operation (pre-failure)
Normal Operation BC1 Failure Operation
BC0 RDM Model
Advertised BC Normal Operation (BC1)
40M Failure mode (BC0) 80M
2
LSP-1(30M) BC1
5
1
4
3
5
Example operation (pre-failure)
Normal Operation BC1 Failure Operation
BC0 RDM Model
Advertised BC Normal Operation (BC1)
40M Failure mode (BC0) 80M
2
LSP-1(30M) BC1
5
1
4
LSP-2(20M) BC1
3
6
Example operation (failure)
Normal Operation BC1 Failure Operation
BC0 RDM Model
Advertised BC Normal Operation (BC1)
40M Failure mode (BC0) 80M
2
LSP-1(30M) BC1
5
1
4
LSP-2(20M) BC1
3
TEAR
7
Example operation (failure)
Normal Operation BC1 Failure Operation
BC0 RDM Model
Advertised BC Normal Operation (BC1)
40M Failure mode (BC0) 80M
2
LSP-1(30M) BC1
5
1
4
LSP-2(20M) BC0
3
8
Example operation (re-optimise)
Normal Operation BC1 Failure Operation
BC0 RDM Model
Advertised BC Normal Operation (BC1)
40M Failure mode (BC0) 80M
2
LSP-1(30M) BC1
5
1
4
LSP-2(20M) BC0
LSP-2 BC1
3
Shared resource reservation crosses BCs
9
Example operation (re-optimise)
Normal Operation BC1 Failure Operation
BC0 RDM Model
Advertised BC Normal Operation (BC1)
40M Failure mode (BC0) 80M
2
LSP-1(30M) BC1
5
1
4
LSP-2(20M) BC1
3
10
Purpose of Draft

• Intended RFC status BCP
• No protocol change required assuming the
  following is possible
• Make-before-break with a new LSPid and a
  different Class Type.
• Bandwidth sharing between different BCs during a
  Make-before-break operation
• Requires common implementation

11
Next Steps

• Feedback from List
• Move to WG Draft