The technology in question is Split Multi-Link Trunking, SMLT, developed by the Toronto, Ontario-based telecoms and data networking vendor in response to the limitations of both the Spanning Tree and Rapid Spanning Tree protocols, said Simon Wilson, Nortel’s Ethernet switching portfolio leader for EMEA.
Dual links, link aggregation, and dual homing are all techniques to add resiliency to the connections between switches, but they also introduce the risk of looping, particularly for broadcast packets. The Spanning Tree protocol was developed to address this issue, which it does by turning off one of the hops in the network. However, this approach reduces link resiliency because it then takes from 30 seconds to a minute-and-a-half to fail over when a link goes down.
An improved version of the protocol, Rapid Spanning Tree, RST, reduces the failover time to sub-second levels, but is in turn vulnerable to loss of the route bridge (the switch holding the network topology information), in which case failover takes up to three seconds while a new route bridge is elected. There have been other attempts to resolve this issue such as Multiple Spanning Tree Groups, MSTG, but this approach is both complex and vulnerable because the static nature of the groups makes load-balancing impossible.
In 2001, Nortel came up with SMLT, which it said had distinct advantages over Spanning Tree, RST, and MSTG. Multi-Link Trunking was Nortel’s name for its proprietary link aggregation technology before the 802.3ad standard existed. SMLT builds on that technology by making two core switches look like a single entity to edge devices, even though they are still operating separately. In this way you get link aggregation, failover with dual homing and load balancing, without the risk of looping, said Wilson.
There are other non-standards-based approaches such as 3Com’s eXpandable Resilient Networking, XRN, and Extreme Networks’ Ethernet Automatic Protection Switching, EAPS, though EAPS is specific to ring topologies, and so is really more of a competitor to the Redundant Packet Ring standard (802.17) than to Spanning Tree. Wilson said Cisco, however, has so far stuck with standards-based approaches. Cisco hasn’t come out with anything so far to address the shortcomings of the other protocols, he said.
Competitors who only offer the standard-based approaches argue that SMLT is a proprietary play, and while Nortel is making efforts to turn it into a standard, Wilson foresees a long process, with ratification probably not before the end of 2007. One problem is which working party to approach within the IEEE, he said. There’s the link resiliency group that works around 802.3ad, but also the network resiliency group, either of which could look at SMLT. Beyond this practical consideration, however, there is the political one. Our competitors clearly have no interest in resolving the standardization issue quickly, said Wilson.
In the meantime, Nortel plans to extend the SMLT capability to switches lower down its portfolio, having until now offered it only on the high end. Most of Nortel’s success and product differentiation in Ethernet switching has traditionally been at the high end, with features such as SMLT, QoS, security, and the delivery as default of extra headroom on the hardware to accommodate future capabilities and technological advances, Wilson said. In the coming year we’ll be extending that differentiation down into lower-cost platforms, with the first announcements in this context in the first quarter.
He is confident of SMLT’s technical superiority over the standards-based alternatives. Cisco only offers RST and MSTG, and in tests against them SMLT was not only better in link and network resiliency, but also simpler, in that it could be set with just 74 commands, whereas the Cisco alternatives took 210, he said.
As to the technology’s applicability to companies other than big enterprises, he said: SMLT is appropriate for any SMB that’s adding resiliency for things like IP telephony.
