PCE+

Background

An ingress runs the CSPF algorithm and uses information stored in the TEDB to calculate MPLS TE tunnels. An MPLS TE network has multiple tunnels. Generally, each tunnel has a different ingress node, and each ingress node performs path calculation independently. As a result, network-wide resource utilization cannot be taken into consideration during path computation, leading to low network resource utilization.

The PCE+ solution can help resolve the preceding issue. This solution involves two device roles: server and client. A server is a path calculation node that stores the path information of the entire network. A client (usually the ingress of a tunnel) is the initiator of a path calculation request. The server computes paths based on the resource usage of the entire network, maximizing network resource utilization.

Benefits

Related Concepts

PCE server

As defined in relevant standards, a PCE server is an entity that can use network topology information to calculate paths or constrained routes. A PCE server can be an application, a network node, or a server. A PCE server on an MPLS TE network receives a calculation request sent by an ingress and uses TEDB information to calculate an optimal constrained path for an MPLS TE LSP.

PCC

A path computation client (PCC), which is usually the ingress of an MPLS TE tunnel, sends path calculation requests to a PCE server.

PCEP

The Path Computation Element Communication Protocol (PCEP), defined in relevant standards, exchanges information between a PCC and a PCE server, as well as between PCE servers in different domains.

Domain

A domain can be an IGP area or a BGP AS. The NetEngine 8000 F supports IGP areas only.

LSP DB

An LSP state database (DB) stores the LSP attributes of an entire MPLS network, which are advertised by PCCs to PCE servers.

Stateful PCE

Stateful PCE is used to construct LSP DBs to monitor LSP information, including the assigned bandwidth and LSP establishment status, and use the LSP DB and TEDB information to compute optimal paths for LSPs on an MPLS network.

Implementation

PCE first performs PCE Discovery. After members are discovered, a PCC and the PCE server establish PCEP Sessions to exchange information. Before the ingress functioning as a PCC establishes an MPLS TE tunnel, the ingress sends a request to the selected PCE server to calculate a path and waits for the calculation result. Unlike IETF PCE, the NetEngine 8000 F supports manual and automatic verification and acceptation of the computation result. After the computed path is confirmed, the PCE server sends the computation result to the PCC. Upon receipt the computation result, the PCC establishes an LSP.

To improve network bandwidth utilization and simplify network operation and maintenance, the NetEngine 8000 F implements Stateful PCE and LSP Re-optimization and Uniform TE Network Information Configuration and Management.

PCE Discovery

Before initiating a path computation request to a PCE server, a PCC needs to discover an available PCE server. The PCE server is passive in path computation and does not need to proactively discover PCCs. The NetEngine 8000 F only supports manually configured PCE member relationships. The source address must be specified on the PCE server. This address is used to set up connections with PCCs. You can specify multiple candidate PCE servers for a PCC. The PCC selects a server based on the priorities and source IP addresses of the servers. If two candidate PCE servers have the same priority, the PCC selects a server with a smaller source IP address. After a server is selected, other servers function as backup ones, and a new server is automatically selected by the PCC if the previously selected server fails.

PCEP Sessions

PCEP sessions need to be established between PCCs and PCE servers, as well as between PCE servers in different domains. After PCEP sessions are established, they can be used to exchange information, including path computation requests and results, between the devices.

Table 1 demonstrates the process of PCEP session establishment, maintenance, and closure.

Table 1 PCEP session implementation

Stage	Diagram	Description
PCEP session establishment	Figure 1 PCEP session establishment	A PCC initiates a TCP request to a PCE server. After the PCC and PCE server perform three handshakes, they establish a TCP connection. The PCC and PCE server exchange PCEP Open messages to negotiate a session. After both devices accept one another's parameters, the two devices exchange Keepalive messages to confirm the negotiated parameters and use these parameters to establish a PCEP session.
PCEP session maintenance	Figure 2 PCEP session maintenance	The PCC and PCE server periodically send Keepalive messages to maintain the PCEP session as long as the session is not closed. The Keepalive message transmissions on the two ends are independent of each other. If one end does not receive the Keepalive message from the other end within a period, the session is considered to be down.
PCEP session disconnection	-	The node, either a PCC or PCE server, that fails to receive a Keepalive message sends a Close message to terminate the PCEP session.

Intra-Domain Path Computation

Intra-domain path computation involves message exchanges only between the PCC and PCE server. In the following example, the TEDB on each node contains the information about the entire network, and a session between the PCC and PCE server has been established. Figure 3 illustrates intra-domain PCE path computation. Table 2 describes inter-domain path computation.

Figure 3 Intra-domain path computation

Stateful PCE and LSP Re-optimization

MPLS TE is expected to properly assign network resources and improve network bandwidth utilization. However, the current TE LSP establishment mechanism cannot perfectly meet these expectations. In Figure 4, each link has 10 Gbit/s bandwidth. The LSP between nodes A and E needs 6 Gbit/s bandwidth, the LSP between nodes C and D needs 8 Gbit/s bandwidth, and the LSP between nodes C and G needs 4 Gbit/s bandwidth. The setup priority of the LSP between nodes A and E is the highest. The C-to-D link has less than 12 Gbit/s bandwidth and a priority lower than the A-to-E link. Without stateful PCE, three LSPs shown in Figure 4 (a) will be established. These established LSPs occupy all links on the network, which is not expected.

Figure 4 LSP establishment with and without stateful PCE

To address this issue, the NetEngine 8000 F implements stateful PCE to improve network bandwidth utilization. After stateful PCE is used, three LSPs shown in Figure 4 (b) will be established. The bandwidth of the links between A and B, B and C, and D and E remain available, optimizing networkwide bandwidth utilization.

Uniform TE Network Information Configuration and Management

Stateful PCE enables a PCE server to uniformly configure and manage TE network topology and tunnel attributes, which facilitates network management and maintenance. The PCE server uses the configured TE topology information and tunnel attributes to compute paths.