GMPLS Overview

GMPLS home

The quick acceptance of Generalized Multi-Protocol Label Switching (GMPLS) by the industry is really a testimonial to the versatility of its predecessor, MPLS which effectively presented a method for two opposing camps, i.e., IP vs. ATM to coexist and establish end-to-end paths in packet-based and cell-based networks to make packet/cell forwarding an efficient operation. Additionally, with the inclusion of traffic engineering extension to signaling protocols, RSVP-TE and/or CR-LDP-TE, it laid the foundations of providing a QoS scheme to IP.

G-MPLS takes the works of MPLS a few step farther, and attempts to consolidate creation of end-to-end Label Switched Paths (LSP) for several types of networks: TDM, and Optical in addition to IP, ATM or Frame Relay.

Polaris has dedicated a separate web page to track the advances and provide useful information on GMPLS. We hope that you find this educational and fun to read at the same time.

GMPLS - What's all the hype about?

Almost every manufacturer of networking equipment has jumped on the bandwagon of GMPLS and claims support for this protocol. So how is it possible that everyone all of a sudden can lay this claim, while the standards draft is still in a state of flux, as of this writing? The answer lies in the fact that GMPLS is basically a superset of MPLS with new additions to address connectivity for TDM and Optical networks. Therefore, if an equipment is currently supporting MPLS, it can claim support for GMPLS, at least partially. At this point in time, we are not aware of a single company that can claim full support for GMPLS and there is very good reason for that: GMPLS covers many equipment types (multiservice edge switches, metro transport equipment, photonic switches) that span many network types (Packet, TDM, Optical) and therefore each network segment, e.g., access, metro, long-haul, would support different parts of the standard draft.

But what is GMPLS? What does it try to solve? Why is it such a hot topic these days? GMPLS for Generalized Multi-Protocol Label Switching is the next generation of MPLS which extends label switching to other media, specifically, time-based (TDM), wave-based (Lambda) and physical-based (Fiber). The basic premise of MPLS is to expedite packet/cell forwarding within the equipment that make up a network segment while providing traffic engineering for the flow itself. A carrier can, then, selectively provide a variety of tiered services for its customer pool. This is fine if all of the network technologies were confined to IP/FR/ATM base.

Covering the past and the near future, GMPLS provides the same enhanced forwarding schemes for TDM, WDM, and physical fiber as it does for an IP/FR/ATM flow. An MPLS "label" is no longer an ID selected randomly from a pool, but is now able to map to a wavelength, optical port or a TDM time slot. The most interesting side-effect for an end-to-end GMPLS capable network is that the entire span can now be effectively automated and the network can become its own manager. Theoretically, based on a given set of requirements, the network, can establish end-to-end connections without requiring the operator(s) to provision every intermediate node. Imagine the time-savings and error-avoidance in that which directly translates into lower cost and potentially higher revenues for the network operator.

GMPLS additional features and requirements

The biggest addition is a new signaling protocol, Link Management Protocol (LMP), to establish, release and manage connections between two adjacent GMPLS-capable nodes. LMP is particularly useful in optical networks as a large pool of Lambdas can be present between two connected nodes which can be problematic using brute force approach in their management. Briefly, the LMP provides for the following:

• Separate data and control channels which can each be protected and accounted for separately o Multiple control channels between two nodes allows for backups and therefore, a graceful switch over from one to another, therefore, maintaining the reliability and the integrity of the network by protecting signaling messages. LMP helps with link-fault localization which is negotiated during the initial "Hello" exchange. o Verification of physical connectivity between two neighboring nodes, which reduces the probability of error in provisioning services.

• Separate data and control channels which can each be protected and accounted for separately

• Multiple control channels between two nodes allows for backups and therefore, a graceful switch over from one to another, therefore, maintaining the reliability and the integrity of the network by protecting signaling messages. LMP helps with link-fault localization which is negotiated during the initial "Hello" exchange.

• Verification of physical connectivity between two neighboring nodes, which reduces the probability of error in provisioning services.

• Extensions to RSVP-TE, or CR-LDP for traffic engineering and establishment of LSPs in optical core networks.
  

• A new, useful feature is allowing for establishment of "bi-directional" LSPs which allow for fast restoration schemes required in non-Packet Switched Capable networks, e.g., TDM.

• Extensions to OSPF-TE and IS-IS-TE to account for optical resources, e.g, fiber id, wavelength bandwidth.

• Hierarchical LSPs, which allow for nested or tunneled set of LSPs of different media, carrying a mix of traffic. This is very helpful when nesting multiple lower-speed LSPs within a single higher-speed LSP. For example, an optical LSP can be established with an OC-192 bandwidth. Nesting numerous STS-1 TDM, or 10/100BT PSC can be accomplished within that single Optical LSP.

• Optimization of internal databases by allowing for Link Aggregation (for LSC)and Unnumbered Links, having no IP addresses (e.g., for TDM links).

Challenges Ahead

Is GMPLS trying to be everything to everyone? Is it complex enough to be shunned by the community when it comes to actual implementation? How about interoperability? The answer to all of these questions lie in the fact that as the optical networking has evolved, it has become more complex as it has to support a variety of traffic mix. Therefore, the burden of complexity is placed either on the equipment manufacturers or their customers. Polaris believes that the customer need not absorb the increasing complexity of the network to provision end-to-end services. But that the more complexity that is hidden from the customer, the easier, more reliable and fail-proof is their operation. And that is where GMPLS will triumph.