What do enterprises want from their network services? The same thing any customer wants of any good or service they purchase -- “I want more for less.” New service features? Sure, as long as they result in lower costs for the customers. New billing models like on-demand pricing? Sure, as long as using them lowers cost. If you talk to enterprise planners, their focus is on making changes that lower their overall costs, period.
What about the providers of those network services? Their universal goal is “I want to stop my profit slide.” If there was ever an example of irreconcilable differences, the gap between what customers want (more for less) and vendors want (more profit) is it. How can enterprises spend less while providers make more? What we need is a revolution, and optical routing just might be it.
Optical networks have done a lot for us. Big fat pipes cost less per unit of capacity than little skinny ones, which means that when you build networks you should quickly concentrate traffic to supersized levels, where the economies of optical advances like wavelength-division and dense wavelength-division multiplexing (WDM and DWDM) can be applied in the core. Optical network vendors like Ciena have been doing pretty well because optical networks are saving money there.
This all may sound like we’ve already had our optical networking revolution, but that’s not the case for two reasons. First, the core isn’t everywhere. In fact, network operators tell me that core networking accounts for less than a quarter of their network costs, and that their greatest cost concentration is at the edge. Even big cost savings in core network can get diluted by other costs, to the point where they create only a dimple on the bottom line. Second, the concentrating of traffic inward from the network edge is almost always done with electrical devices in what’s often called the “aggregation layer”. And any time you say “layer” you’re layering on capex to buy things like routers, and opex to manage them. All those costs erode economy of scale benefits
Suppose you could push optics out closer to that costly edge by making the router a part of your optical network. This is where optical routing comes in. Today, routers use optical Ethernet interfaces that aren’t compatible with the dense wavelength-division multiplexing (DWDM) used in modern optical transport networks. However, if we substituted DWDM interfaces for optical Ethernet, we could attach routers right to our optical core, with no additional layers, their costs, and their management burdens.
With DWDM interfaces on routers, it’s possible to build optical networks that switch wavelengths (using “reconfigurable optical add-drop multiplexers” or ROADMs) that reach way out toward the edge. These increase capacity at the edge, reducing the congestion that creates capacity management challenges. The optical transport standards also include forward error correction (FEC) so that bit errors on a path are corrected in flight and don’t create a need for retransmission. That also reduces management burdens, because retransmission ties up devices. Finally, optical handling with fewer electro-optical conversion layers means lower latency, and that means delay-sensitive applications can be supported better.
Optical routing could create a vast reservoir of capacity that reaches even the user. Enterprises use multiprotocol label switching (MPLS) VPNs for their major site connections, and these have an access router at the end of an optical path. If that router had a DWDM-compatible interface, we could push a full wavelength to every MPLS VPN site. We could then use a bigger “metro” router to terminate all these wavelengths, which is more economical than having smaller per-customer routers in that mission. The cost of this should be a bit lower than we see today, and the capacity potential should be a lot higher.
The packet optical innovation, fully exploited, could actually make our enterprise’s “more for less” wish a reality, a transformational shift for the telcos in general, and for enterprise service profits. It could also perhaps limit how many current VPN sites are displaced by low-cost, Internet-based, SD-WAN services. That could protect more of the MPLS VPN revenues operators are now at risk of losing, and that could help justify investing in optical routing.
The final benefit of optical routing is the overall improvement in latency we see from the combination of higher capacity and fewer devices in the data path. Many IoT applications require low latency, and today they’re supported on computers placed close to the IoT elements because network latency could compromise the accuracy of the control systems they run. Could the lower latency optical routing provides create an opportunity to pull “the edge” into the network, creating a new service opportunity for telcos and cloud providers? Yes, it could.
Optical routing could be our needed revolution, but for that to happen we need more confidence in the strategy on both sides. Only a third of enterprises I’ve talked with believe that the telcos will be able to make network changes that will allow them to hold the line on costs, and less than five percent think that more-for-less will be offered them. Interestingly, more telcos believe in optical routing than enterprises; half those I’ve talked with think it will reduce network costs “significantly” and just under 20% think it could actually let them offer improved performance with at least a slight reduction in price. If optical routing is potentially this transformational, why aren’t telcos all over it?
It may be the equipment vendors. Even though DWDM interfaces on routers would create incremental revenue for router vendors, it might also let DWDM/ROADM optical networks undermine core router sales in the near term, and for stock-price-fixated vendor executives, that’s a bad thing. Half of telcos say that vendors aren’t pushing optical routing, only offering it if asked. We’ll have to wait and see whether benefits to both telcos and enterprise service buyers can overcome vendor obsessions with making their quarterly numbers.