Traffic grooming in optical networks -- sometimes known somewhat inappropriately as optical traffic grooming or marketed as sub-lambda traffic grooming -- is the technology that allows network operators to transport
This problem was solved traditionally with digital cross-connects (DXCs) or other traffic grooming network elements that combined constant bit rate (CBR) streams into a higher-speed container/frame/envelope, using either SONET/SDH framing or Generic Framing Procedure (GFP; ITU-T specification G.7041). These devices provided time division multiplexing (TDM), giving each subrate bit stream a fixed number of slots in a container transported in a higher-speed bit stream.
The all-optical solutions (including all-optical add-drop multiplexers) that became available in the last few years have reduced the complexity of high-speed optical networks by reducing the number of optical-electrical-optical conversions in the network. Unfortunately, no such advances have been made in traffic grooming technologies. No all-optical traffic grooming products are available on the market, so all traffic grooming is still performed in the electrical domain, as it was in the days when the SONET/SDH transmission technology dominated service provider networks.
Optimizing optical networks without optical traffic grooming aids
With all these facts in mind, one has to wonder about the vendor announcements touting concepts from "lambda-level grooming" to "subport-level grooming." These are just marketing terms for a simple fact: Since most traffic has been migrated to IP, it makes no sense to build new networks with TDM grooming equipment. You should optimize your backbone to include only the elements relevant to your services: the switching platform (routers, also known as Layer 3 switches) and the transport platform (WDM).
Real-world dilemmas are always more complex than long-term architectural visions. You should never blindly follow a generic architecture but instead compare the total cost of ownership (TCO) of various alternatives to select the best architecture for your network. The TCO should obviously include acquisition costs and the costs of operating, maintaining and monitoring multiple platforms, potentially acquired from multiple vendors and using different network management systems.
Analyze the services traffic on your network. For example, if you have existing services on your TDM grooming equipment (e.g., voice, Frame Relay or ATM traffic), don't migrate them to IP just to build a next-generation converged backbone unless the cost-benefit analysis clearly shows a short-term return on your investment. Usually, it's better to keep the old TDM equipment operational as long as you provide the "old-world" services. Don't even try to integrate it with your new network; just give it a separate set of wavelengths on your WDM network.
Know when to eliminate network layers. On the other hand, if you're expanding your IP network, try to eliminate as many layers as possible. You should deploy routers and switches connected directly to the optical network. Ideally, these devices would already have DWDM-compliant optical interfaces, eliminating the transponders -- another costly part of optical networks that performs optical-electrical-optical conversion between multimode fiber interfaces and DWDM wavelengths. Buying new SONET/SDH gear to connect your routers to the WDM core no longer makes sense.
Note on borderline cases. If, for example, you need to transport two 1 Gbps streams but lack high-speed DWDM interfaces on your Layer 3 gear and free wavelengths in your optical network, use the grooming functionality available in many DWDM transponders. The transponder card will be cheaper than a full-blown DXC.
Replicating SONET functionality in IP-over-DWDM networks
When designing your next-generation network, don't forget that the brave new world of IP-over-DWDM networking does not necessarily provide the same functionality as the SONET/SDH world. You might lack the enhanced error detection and reporting provided by SONET/SDH unless the DWDM interfaces of your routers support Optical Transport Network (OTN) framing.
Nor is fast failure detection and recovery a given in a router-only network. You have to use the Fast Reroute functionality of MPLS Traffic Engineering (MPLS TE) to reach rerouting times comparable to SONET/SDH. And don't forget Quality of Service (QoS). When using TDM gear, it's obvious how much bandwidth each connection will get; it cannot get any more or less bandwidth than what you've provisioned. That might not always be the case in IP/MPLS-based backbones unless you consider the QoS implications in your network design and implementation.
Finally, using Layer 3 devices (routers or switches) to perform transit traffic switching in an optical ring is expensive. Keep as much transit traffic as possible in the all-optical domain, and try to use routers as you would use add-drop multiplexers in traditional TDM architectures.
About the author: Ivan Pepelnjak, CCIE No. 1354, is a 25-year veteran of the networking industry. He has more than 10 years of experience in designing, installing, troubleshooting and operating large service provider and enterprise WAN and LAN networks and is currently chief technology advisor at NIL Data Communications, focusing on advanced IP-based networks and Web technologies. His books include MPLS and VPN Architectures and EIGRP Network Design. Check out his IOS Hints blog. He is also a SearchTelecom.com expert. Ask him your networking questions.
This was first published in April 2010