MPLS - network redundancy
When implementing an MPLS network, ISPs will typically design a network that has built-in redundancy, resiliency and load balancing to maximize efficiency and availability. There are several techniques used to achieve this, but all of them depend first and foremost on a network with many P routers, and multiple alternate paths.
MPLS, traffic is forwarded through Label Switched Paths (LSPs) which are established between ingress and egress routers (PEs). You can set up multiple LSPs between the same ingress and egress routers to achieve load balancing and redundancy across multiple paths. This can be achieved in one of several ways:
- Equal-Cost Multi-Path (ECMP): When multiple paths with equal cost (i.e., equal IGP metric) exist between the ingress and egress routers, the traffic is load balanced across these paths. The traffic is typically divided based on a hashing algorithm that considers parameters such as source and destination IP addresses, protocol, and port numbers. ECMP can also be used in conjunction with MPLS Traffic Engineering (TE) to optimize resource utilization and achieve load balancing.
- MPLS traffic engineering (TE): MPLS TE allows network operators to establish LSPs with specific constraints, such as bandwidth or latency requirements. Using MPLS TE, multiple LSPs with diverse paths can be created to distribute traffic across the network effectively. Load balancing in MPLS TE can be achieved by assigning different weights to the LSPs or by using dynamic load-balancing mechanisms, such as flow-based or adaptive load balancing.
- LSP Multipathing: In LSP multipathing, multiple LSPs are set up between the same ingress and egress routers using different paths. The ingress router can then distribute the traffic across these LSPs based on various methods, such as per-flow or per-packet load balancing. Per-flow load balancing is preferred because it helps maintain the correct packet order for individual flows, while per-packet load balancing can result in packet reordering and higher latency.
Any MPLS topology that enables multiple paths within its infrastructure is capable of using any of the above. All MPLS infrastructure is composed of dozens and sometimes hundreds of routers, including many P routers which deliver a multitude of redundant paths that ultimately, with the techniques described above, deliver both robustness, redundancy, and efficiency.