What are the trade-offs between centralized and distributed PCE architectures for MPLS VPNs?

1 Centralized PCE

A centralized PCE architecture consists of one or more PCE servers that are responsible for computing the paths for all the MPLS VPNs in the network. The PCE servers communicate with the network devices using a protocol such as BGP-LS, which provides the topology and link state information of the network. The advantages of a centralized PCE are that it simplifies the management and configuration of the network, provides a global view of the network state and resources, and enables consistent and coordinated path computation across multiple domains. However, a centralized PCE also has some drawbacks, such as scalability issues, single point of failure risks, and potential performance degradation due to network congestion or PCE overload.

2 Distributed PCE

A distributed PCE architecture consists of multiple PCE servers that are distributed across the network and cooperate with each other to compute the paths for the MPLS VPNs. The PCE servers can be either embedded in the network devices or deployed as separate entities. The PCE servers use a protocol such as PCEP, which allows them to exchange path computation requests and responses, as well as synchronization and coordination messages. The advantages of a distributed PCE are that it improves the scalability and reliability of the network, reduces the dependency on a single PCE server, and enhances the responsiveness and adaptability of the path computation to the network dynamics. However, a distributed PCE also has some challenges, such as complexity and overhead of the inter-PCE communication, inconsistency and conflicts of the path computation results, and difficulty of enforcing global policies and constraints.

3 Hybrid PCE

A hybrid PCE architecture is a combination of centralized and distributed PCEs, where some PCE servers act as master or leader PCEs, and others act as slave or follower PCEs. The master PCEs have a global view of the network and can delegate some of the path computation tasks to the slave PCEs, which have a local view of the network and can perform faster and more efficient path computation. The hybrid PCE architecture aims to balance the benefits and drawbacks of both centralized and distributed PCEs, by providing a hierarchical and flexible structure that can adapt to different network scenarios and requirements.

4 Factors to consider

When selecting between centralized, distributed, or hybrid PCE architectures for MPLS VPNs, there are a few factors to consider. These include the size and complexity of the network, the traffic engineering objectives and constraints, the network topology and connectivity, the network dynamics and variability, and the network management and operation. Depending on these factors, different PCE architectures may offer different levels of performance, efficiency, robustness, and flexibility for MPLS VPNs.

5 Best practices

Optimizing the PCE architecture for MPLS VPNs requires following certain best practices. This includes defining traffic engineering policies and objectives that align with business goals and SLAs, choosing the right PCE protocol and mode for the network devices and PCE servers, designing and deploying PCE servers with sufficient capacity, redundancy, security, and resilience, evaluating and comparing the trade-offs between centralized, distributed, and hybrid PCE architectures, and testing and validating the PCE architecture. By doing this, you can leverage PCE to improve traffic engineering and path computation for your MPLS VPNs.