Is PCEP/BGP-LS-based SDN Approach Ideal Choice for Service Providers?
Author: Kirit Bhamre.
The Software Defined Networking (SDN) movement offers greater benefits to make networks more customized, efficient, application centric, and programmable. There are many different approaches to build SDN. OpenFlow and VXLAN protocols are gaining more traction in data center environments, while Segment Routing and PCEP/BGP-LS are carriers’ SDN technology option for obvious reasons. Unlike OpenFlow/VXLAN Segment Routing and PCEP/BGP-LS doesn’t require forklifting their existing network, these protocols based solution can be implemented by doing Software upgrade. The migration path to PCE-based SDN is evolutionary, with lower CAPEX and OPEX than alternative approaches.
In this blog, I would like to focus on Path Computation Elements Protocol (PCEP), and how it can help enabling secure, dynamic, optimal, and inter-area as well as inter-domain traffic engineered path setup.
What is PCE?
Path Computation Elements (PCE) is an entity that computes paths based on supplied constraints on behalf of routers, an OSS, or another PCE in the network. When a network node needs a path for an LSP (labeled switch path), it makes a request to the PCE using PCE protocol (PCEP). The PCE has the access to topology information for the entire network domain and uses this information for patch computations. The PCE architecture and PCE protocol is defined by IETF in RFCs 4655 and 5440 respectively.
How does PCEP works?
The session between Path Computation Client (PCC) and Path Computation Element (PCE) gets established over TCP (just like BGP). Once the session is established, PCE builds topology database (TED) using underline OSPF/ISIS or BGP-LS. In OSPF/ISIS and BGP-LS has special TLV capabilities have been added for PCE to learn/build this database. When PCC requests the LSP with certain constraints (i.e. bandwidth, cost, priority, etc…), PCE computes the path based on topology database and respond with appropriate path information.
This centralized approach to maintaining the topology database and performing path computing/provisioning helps increase network performance, adds flexibility, and better use of network resources.
The main purpose of PCE is to solve multi-domain problem. PCE should be able to have topology visibility not only in single domain, but also build a topology database across multiple domains.
There are four methods have been defined to build topology database across multi-domain:
- Per domain path computation
- Simple co-operating PCEs
- Backward Recursive Path computation
- Hierarchical PCE
The following diagram shows how co-operating PCE helps provisioning LSP across multiple domains.
A PCE-based SDN approach is an ideal solution for service providers. SDN’s primary goal is to decouple the control plane from data plane, and have a centralized control plane mechanism that takes control for path provisioning. PCEs provide centralized control of paths that are setup for flows in MPLS networks. Deploying a PCE, with appropriate policy to the OSS, is the quickest and easiest way to achieve SDN in traditional Telco/Transport network. By deploying PCEs, service provider’s MPLS-based network can gain substantial benefits of inter-domain routing, customizable path computation, increased network performance, and better network resources utilization.