TL;DR
This paper proposes an experimental procedure and extensions to the Path Computation Element Protocol (PCEP) for computing Point-to-Multipoint (P2MP) Traffic Engineering Label Switched Paths (TE LSPs) across multiple network domains. This is crucial for enabling scalable multicast services like IPTV and content distribution over complex, multi-domain networks.
What Problem Does This Paper Address?
Modern networks are increasingly supporting high-capacity multicast services, such as IPTV and content delivery networks. These services require efficient routing of data to multiple destinations simultaneously, which is achieved through Point-to-Multipoint (P2MP) Traffic Engineering Label Switched Paths (TE LSPs). However, computing these P2MP TE LSPs across multiple administrative domains (e.g., different service providers) presents a significant challenge. Existing methods for inter-domain path computation, often based on point-to-point (P2P) approaches, struggle to efficiently and optimally construct P2MP trees that span multiple domains. This difficulty arises from the need to coordinate path segments across domain boundaries, manage potentially complex branching structures, and maintain domain confidentiality, all while aiming for optimal tree cost and stability.
Methods
The paper introduces a "core-tree" based approach for inter-domain P2MP path computation. This method breaks down the complex P2MP tree construction into two main steps:
- Core-Tree Computation: This involves computing a "core-tree" that spans the domains. The root of this core-tree is the ingress router in the source domain, and its leaves are the entry boundary nodes of the destination domains. This core-tree represents the inter-domain backbone of the P2MP LSP. The paper suggests using an extended Backward-Recursive PCE-Based Computation (BRPC) procedure for this step.
- Sub-tree Grafting: Once the core-tree is established, sub-trees are computed within each leaf domain and then "grafted" onto the core-tree. This allows for the efficient computation of the final P2MP LSP from the ingress LSR to all egress LSRs.
To support this experimental procedure, the paper proposes extensions to the Path Computation Element Protocol (PCEP). Specifically, it introduces a new "C-bit" (Core-Tree bit) in the Request Parameters (RP) object. This bit signals whether a PCEP message is related to inter-domain P2MP core-tree computation or sub-tree grafting.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- Decomposition for Scalability: Breaking down the complex inter-domain P2MP path computation into a core-tree and sub-tree phases significantly simplifies the problem and improves scalability.
- Core-Tree as a Backbone: The core-tree acts as a stable inter-domain backbone, allowing for more localized sub-tree computations within individual domains. This reduces the complexity of coordinating path segments across multiple administrative boundaries.
- PCEP Extension for Inter-Domain P2MP: The introduction of the C-bit in the PCEP RP object provides a clear mechanism for signaling and differentiating inter-domain P2MP core-tree operations from other PCEP messages.
- Confidentiality Support: The proposed approach can leverage mechanisms like Path-Keys to preserve topology confidentiality between domains, which is crucial for inter-provider networks.
- Trade-offs in Computation: The paper acknowledges that optimal P2MP tree computation can be NP-complete. The core-tree approach allows for trade-offs between optimality, computation time, and stability, potentially enabling offline computation of the core-tree for better stability.
Key Results
The paper doesn't present quantitative results from experimental deployments as it's an experimental protocol. However, it outlines the proposed procedures and PCEP extensions that are designed to:
- Enable the computation of inter-domain P2MP TE LSPs.
- Provide a more scalable and manageable approach compared to trying to compute the entire P2MP tree end-to-end across domains.
- Support confidentiality requirements in multi-domain environments.
- Allow for flexibility in optimization objectives and constraints.
Limitations and Future Work
- Scalability of Dynamic Computation: The paper notes that the dynamic computation of the core-tree, especially with iterative checking of all combinations, might present a scalability problem for dense or meshed domains.
- Offline vs. Online Computation: While the core-tree can be computed offline for stability, the paper also discusses opportunities for parallel online computation, highlighting a trade-off.
- Future Work: The document states that when implementation and deployment results are available, the document will be updated and refined, and then moved from Experimental status to Standards Track. The algorithms for computing optimal large sub-trees are also explicitly stated as being outside the scope of this document, suggesting this as a potential area for future research.
Summary updated