PCE requirements and architecture
The control of MPLS or GMPLS network demands efficient path computation engine to provide paths that:
- Fulfil service level agreement (i.e. QOS, protection scheme).
- Optimise network resource usage in a complex ( multilayer and multi domain) and large network ( multi ISP with partial visibility)
In the MPLS and GMPLS architecture the path computation was typically performed inside network elements. However, this approach has shown some limitation due to:
- High complexity of specific application such as point to multipoint or inter-layer path computation.
- Network topology lack of visibility and, confidentiality issue in inter-carrier case.
That is why IETF has a proposed a Path Computation Element Based Architecture as defined in RFC4655.
It defines the use of a new functional equipment called PCE in charge to compute the path on any client request using complex algorithm and Traffic engineering information collected through routing protocols.
Composite PCE versus an External PCE
This architecture is mainly formed of 3 types of protocols:
- The PCE communication protocol (PCEP) which is a TCP client/server application where a client (PCC) can request a path to a path computation server (PCE).
- A discovery protocol to help PCC to know about existing PCEs.
- An IGP that enables the PCE to get the traffic engineering database of the whole Networks.
The PCE protocol that has been elected is a new IP protocol acting as a client server protocol with extensions that enable cooperation between PCE to solve inter domain routing.
Find below the scheme for such.
Protocol exchange through 3 different PCEs which cooperate to build the explicit Route
