MPLS Control Plane for Packet Networks
The original intent of MPLS was to increase the speed of IP forwarding and to give
traditional IP architectures (which at the time were "best effort delivery"
with no proper QoS) the connection-oriented behavior of alternative networks such as ATM.
It didn’t take long for MPLS equipment to surpass these goals.
The early MPLS deployments were concentrated in routers. Transport was therefore
over packet networks. These routers could offer a number of services, for example
BGP/MPLS Layer 3 VPNs, over a single network and such network convergence was
one of the early drivers of MPLS deployments. However, an MPLS control plane
provides a number of other benefits for IP/Packet networks: traffic engineering,
guaranteed QoS, fast protection and restoration and OAM.
Traffic Engineering (TE)
Traffic engineering is the process of steering traffic across a network
to optimize network performance and resources. Pure IP networks do not offer
this capability, normally utilizing alternative routes only when the primary route
has failed. Traffic Engineering allows network resources to be used more efficiently.
Traffic Engineering in MPLS is achieved by extending the link state routing protocols,
such as OSPF or IS-IS, to propagate TE-related information through the network.
A Constrained Shortest Path First (CSPF) algorithm is used to select the optimal path
through the network that meets the TE requirements. RSVP-TE is then used to set up
that path through the network.
One particular use of TE is to establish backup LSPs that use diverse links
and network elements from the primary LSP. This means there is no single point of failure.
The TE extensions include Shared Risk Link Group (SRLG) information which not only
guarantees diverse links, but also links that do not share the same fiber bundle or conduit.
Guaranteed QoS
MPLS further builds on the Traffic Engineering capabilities by specifying
Quality of Service (QoS) attributes. At the simplest level this might be the bandwidth
required for an LSP. The TE routing described above routes the LSP through links
that have sufficient capacity. Additional constraints can also be specified if required,
for example delay or jitter.
The RSVP-TE signaling carries the QoS parameters, enabling each network element
along the LSP to reserve any resources needed to meet the QoS requirements. Typically,
each network element would also re-advertise reduced available bandwidth on the links
that have been used. However, statistical multiplexing schemes can be used to allow
over-subscription as needed.
Fast Protection and Restoration
Traditional Layer 2 and IP forwarding typically takes a number of seconds to recover
from a network failure. With the critical nature of much network traffic today,
this is unacceptable. The TE capabilities of MPLS allow alternative backup LSPs
to be pre-provisioned, and for restoration to happen in a small number of milliseconds.
One possibility is to have a complete backup LSP end to end, using a diverse route.
The backup can be carrying the same traffic as the primary (1+1) or be ready for
a quick switch over (1:1 or 1:N). Such "Protection Switching" can often be achieved
with zero or minimal signaling.
DC-MPLS also supports MPLS Fast Reroute per RFC4090. This describes two methods,
both of which are supported:
- 1 to 1 back up LSP (Detour LSP)
- Facility method (Bypass LSP)
DC-MPLS also supports end-to-end (e2e) recovery, where the entire LSP from ingress
to egress is protected. DC-MPLS supports all flavors of e2e LSP recovery per RFC4872:
- Full LSP rerouting (before and after failure rerouting)
- Rerouting without extra traffic (includes shared-mesh restoration case)
- 1:N protection with extra traffic
- 1+1 unidirectional protection
- 1+1 bidirectional protection
OAM
DC-MPLS also supports extensions to the familiar "ping" and "traceroute"
operation and management schemes. These facilitate the detection and localization of errors.
Related links:
For inquiries about Metaswitch's
MPLS products and expertise contact
.