Continuing my exploration of next-generation Ethernet encapsulation technologies, in this post I look at MPLS-TP, TRILL and Provider Link State Bridging.  These technologies are all currently under development in ITU-T, IETF and IEEE.

MPLS-TP

Originally begun in ITU-T as T-MPLS, this project is now incorporated into the MPLS work of the IETF, and has become a set of extensions and simplifications to MPLS.  Multi-Protocol Label Switching is a well established, successful (but ageing) core network technology, and forms the basis of almost every major Carrier Ethernet transport network worldwide.

As with traditional MPLS, the technology involves encapsulating a customer’s Ethernet frames (in-cluding the source and destination addresses, but discarding the FCS) within an LSP and/or Pseudowire at a Provider Edge device.  This is done by adding a label to the front of the incoming frames. Within the network core, these labelled packets are transported hop-by-hop, potentially swapping the label at each intermediate device.  Where the transport links between hops are themselves Ethernet, an additional Ethernet header is added at each device, and removed at the next.

The definition of MPLS-TP is work in progress, and it remains to be seen exactly what it will mean for the Carrier Ethernet industry.  However it may form the basis of a separation of MPLS into different layers, with MPLS-TP forming a simpler transport layer below MPLS service features.  The main new technical aspects of MPLS-TP are the removal of dependence on IP and layer-3 addressing, and improved OAM.

MPLS-TP shows promise as a future direction for Ethernet in transport networks, but its development is dogged by a lack of common goals amongst the participants and political disagreements.  So what's new? Standards developments are not renowned for a lack of disagreement.  The issue is whether the participants genuinely agree on the aims and goals of the technology.

TRILL

There are several current standards projects which aim to remove restrictions and inefficiencies from bridged networks based on spanning trees, including GELS (G-MPLS for Ethernet Label Switching), PLSB (Provider Link State Bridging) and TRILL (Transparent Interconnection of Lots of Links).

TRILL is being developed by an IETF working group and some drafts have reached the WG draft stage.  TRILL replaces spanning tree with a link-state routing protocol (IS-IS) and incorporates a hop-count to allow looping packets to be detected.  Using a link-state protocol allows Rbridges to select the shortest route to each destination rather than sending packets along a shared spanning tree, which achieves better network utilisation in a meshed network.  It is the addition of a hop-count which most offends die-hard layer-2 adherents.  TRILL Rbridges terminate spanning tree frames and thus cannot be used to transparently replace individual links in a bridged network.  The data plane is not the same as that used in 802.1 bridges.

Because of its differences from deployed Ethernet technologies, TRILL may meet resistance in existing markets.  However in “green-field” deployments, it makes an interesting candidate replacement for existing bridging technologies.  One area where TRILL could find a niche is in data centres.

Provider Link State Bridging

P802.1aq is a project to develop Shortest Path Bridging technologies.  The project has similar aims to TRILL but achieves them by extending existing bridging techniques rather than developing a new data plane.  In essence, the idea is to form multiple spanning trees rooted at each data source in a network.  The trees are formed using a link state routing protocol (IS-IS), and are used to send traffic from that source.  Consequently, the shortest path can be used for each datastream.

Changes in the industry, in particular redundancies and takeovers, have had a major impact on the development of PLSB.  In particular, protocol architects who have remained hidden for decades in the R&D labs of large corporations have emerged into the standards community in search of new opportunities.  This has led to some of the most interesting interactions I have seen for years, and it's quite possible that the excitement being seen in IEEE 802.1 over the development of genuinely new ideas may result in real technological advances and give PLSB an edge over the competition.  PLSB is an interesting candidate technology for access and metro core Carrier Ethernet networks.