Transcript Generalized Multiprotocol label Switching: An overview of

(Slide set by
Norvald Stol/Steinar Bjørnstad
2013)
Outline
• Introduction
• Enhancements to signaling
- Hierarchical LSP setup
- The suggested label
- Bidirectional LSP setup
- Notify messages
• GMPLS protection and Restoration
techniques
- Protection mechanisms
(Span/Path protection)
- Restoration mechanisms
• Conclusions
Introduction
• IP -> MPLS
=> Datagram to Virtual Connection (VC)
(point-to-point)
• Explicitly routed label switched paths (LSPs)
established before information transport –
independent of actual routing paradigm
• Label swapping used as forwarding paradigm
• Forwarding equivalence classes (FECs)
• Label hierarchy / Label stacking
Introduction (2)
• Constraint based routing
- traffic engineering (QoS differentiation)
- fast reroute (after failure)
- diversity routing (disjoint alternative paths for
protection)
• Routing protocols (e.g. OSPF) must exchange
sufficient information for ”constraint”
• Resource reservation protocol with traffic
engineering (RSVP-TE) is used to establish
LSP/label forwarding states along path.
(The alternative CR-LDP is not used any more)
Introduction (3)
Generalized MPLS:
• Extensions to handle e.g. optical network
resources (OXC’s) (e.g. extensions of OSPF,
RSVP-TE).
• Common control plane for packet and optical
network
• New Link Management Protocol (LMP) for
optical links.
• Support for (label) switching in time, wavelength
and space domains – and a label hierarchy.
• Additional functionality to handle bidirectional
links and protection/restoration.
RSVP-TE and OSPF enhancements
• RSVP-TE (CR-LDP)
– Initiate optical channel trails
– For optical networks and other connection
oriented networks
• OSPF (IS-IS)
– Advertise availability of resources
– Bandwidth of wavelengths
– Interface types
– Other network attributes and constraints
Enhancements to signaling
• Control plane may be physically diverse from the
data plane.
• Hierarchical LSPs 
(Study the example in the article to see what establishing
a new LSP may entail, start with LSP1)
• The suggested label:
– An upstream node suggests an optimal label (fast)
• May be overridden by its downstream node (slower)
- In optical networks with limited wavelength conversion
– Suggested wavelength (-label) to use is very useful
Enhancements to signaling (2)
Bidirectional LSP setup (New in GMPLS):
• Bidirectional optical LSPs (lightpaths) are important for
network operators
– Fate sharing
– Protection and restoration
– Same QoS in both directions, same resource demands
Problems with two independent LSPs in MPLS:
• Additional delay in set-up (problem in protection)
• Race conditions for scarce resources => lower probability of
success for both directions simultaneously
• Twice the control overhead
In GMPLS:
Single set of Path/Request and Resv/Mapping messages used to
establish LSPs in both directions at once.
Enhancements to signaling (3)
Notify messages:
• Added to RSVP-TE for GMPLS
• Provides a mechanism for informing
nonadjacent nodes of LSP-related failures.
– Inform nodes responsible for restoring
connection
– Avoid processing in intermediate nodes
• Speed up
– Failure detection and reaction
– Re-establishment of normal operation
GMPLS Protection and Restoration
Four primary steps of fault management:
• Detection
- should be handled at layer closest to failure, i.e. optical
layer. E.g. ”Loss-of-light” (LOL), Bit Error Ratio, ..
• Localization
- requires communication between nodes. LMP includes
procedure for fault localization.
– Channel fail message over separate control channel
• Notification
- Notify message added to RSVP-TE signaling
• Mitigation
– “Repairing the failure”
GMPLS Protection and Restoration (2)
• Path switching (End-to-end)
– Failures addressed at path end-points
• Line switching (local)
– Action at intermediate transit nodes where the failure is detected
• Prot and rest. Terms not precisely defined: In practice
used for fault handling in different time frames.
• ”Protection”
– Fast
– usually pre-allocated resources to handle failures quickly,
– e.g. SDH/SONET: 50 ms – 100% extra resources and
simultaneous transmission. (1+1 protection)
GMPLS Restoration
• When fault is handled after a failure has occurred
– Dynamic resource allocation
• Usually at least one order of magnitude higher delay
than protection
• Different levels of ”preparedness”
– Pre-calculated routes or not;
– Some resources reserved or not
GMPLS Protection and Restoration (3)
Protection mechanisms:
• 1+1 protection: simultaneous transmission of data on
two different paths.
• M:N protection: M preallocated back-up paths shared by
N connections. (1:N is most usual; 1:1 also relevant).
• Span protection – between adjacent nodes (NB! Avoid
”fate sharing”):
GMPLS Protection and Restoration (4)
• 1+1 Path protection (disjoint paths):
• For M:N Path protection: back-up paths may be used for
lower priority traffic in normal operation – preemption
(Supported by GMPLS)
GMPLS Protection and Restoration (5)
• Restoration mechanisms:
• Alternative paths may be computed beforehand, but
resources are seldom allocated before they are needed.
Conclusion
• GMPLS is a good idea and do have a lot
of nice functionality to handle the networks
of the future!