The term "convergence" may be one of the most overused and overhyped words in telecoms, but there's no better way to describe the current interest in flattening the IP and optical layers of the network.
The idea of packet-optical convergence - which in broad terms means taking packet networks (namely Carrier Ethernet), Sonet/SDH and DWDM and flattening them down into one network that does everything those layers do separately - has been around for some time.
However, in early 2009, US operator Verizon threw the gauntlet down to vendors when Stuart Elby, VP of network architecture at Verizon Network & Technology, said at an OFC conference that it intended to transform its global network into a packet-optical transport system (P-OTS) that would combine Layer 1 and Layer 2 functionality and into a much more efficient and cost-effective network with an integrated control plane. And Verizon wanted suppliers to come up with boxes that would help them achieve it.
The basis for Verizon's demand was an internal analysis that found IP transit traffic patterns and demand for flexible routes were so dynamic that IP traffic at the optical layer often didn't have to touch the network routers.
"That adage of 'switch where you can, route where you must' has never gone away, and Verizon wanted its IP transit traffic that didn't need routing to stay in the optical layer," explains Anup Changaroth, product marketing director for Asia for Nortel Networks' MEN business recently purchased by Ciena. "It's a very costly affair to put routers in place and take your IP traffic up to that layer if you don't need to."
Verizon concluded that to support those dynamic traffic patterns and bypass routers, "it made more sense to have the optical layer using MPLS-TP as the key switching mechanism. And they've been driving vendors to look at that," says Changaroth, adding that carriers in Japan and elsewhere have done their own internal studies and reached the same conclusions in the last six months.
Infonetics, meanwhile, found in a survey last year that two-thirds of service providers plan to combine their data and transport operations sometime next year. And vendors are now jockeying for position to help them do just that.
We have the technology
The technological advances enabling the push to packet-optical convergence are already here: Ethernet-over-SDH, ROADM (for wavelength-switching), ASON (Automatically Switched Optical Network), GMPLS (which allows MPLS to run on the control plane) and OTN (Optical Transport Network) switching.
OTN is one of the key technologies mentioned in Verizon's P-OTS strategy. Verizon intends to implement a "wavelength-centric OTN-compliant network" supporting multi-vendor interoperable OTN-compliant (G.709) interfaces.
"OTN is key because it brings a lot of the good manageability stuff from Sonet/SDH to optical, so you can see the traffic, detect faults, all the operational management stuff and granularity from SDH," Changaroth says.
It also supports legacy TDM traffic, which is crucial to packet-optical convergence, he adds. "TDM may not be growing by leaps and bounds as much as IP traffic, but it still generates a huge amount of revenue for telcos, so anyone who says they can just get rid of that TDM Sonet/SDH layer is kidding themselves."
Bandwidth and cost efficiencies
Of all the benefits of flattening the IP and optical layers, there are two recurring themes: more efficient bandwidth usage and lower costs.
"By bringing several layers of their network together, service providers can reduce the number of devices in the network, the space and power consumption," says Luc Ceuppens, marketing VP of high-end systems for Juniper Networks. "That will help them not only get capex down but prepare for the future services they want to run over these networks."
Ronen Mikdashi, AVP and head of the product marketing department at ECI Telecom, agrees that packet-optical convergence will help reduce capex and opex, but adds that the savings don't just come from the hardware.
"In this type of convergence, software should support all the layers in a single management system to support end-to-end provisioning, which reduces time to market, response time to customer needs for expansion, and other things that can also help to reduce costs," he says.
Meanwhile, bandwidth efficiency gains are a matter of having the flexibility and sufficient granularity to fill lightwaves to capacity, Mikdashi says.
"If you have a 40G WDM channel and you're only using 10 Gbps of it because your Ethernet service only runs at 10G, you can multiplex several other low-bit subservices onto it, running as low as 2 Mbps," he says. "You can fill it with 4 x 2.5G or 10 x 1G or 100 x 100 Mbps and so on - any combination you like, so all your channels are fully utilized."
Alcatel-Lucent - which fired the first shot in the packet-optical convergence wars with its converged backbone transformation (CBT) strategy launched in September - touted its ability to groom traffic not only at the wave level, but also the sub-port level using ODUflex technology, an emerging ITU standard due for completion next year, which provides higher granularity by enabling VLANs or pseudowires within a port to be logically or virtually mapped to the same wave.
Result: carriers can maximize capacity without spending more money on extra core routers, and yield capex savings of "at least 30%, in addition to savings in power, space and operational complexity".
Optical-centric or packet-centric
A minor war is already brewing over just how much money telcos can save depending on whose solution they choose - or rather, how focused their convergence strategy is on bringing the packet layer to DWDM or the other way around.
There is a difference, says Ceuppens of Juniper - and one that tends to be defined by the core expertise of the vendor.
"You have players in the optical world and the packet world and they'll each approach this new product in their own way," he says. "Optical-centric vendors will develop optical transport with some packet capabilities, and packet-centric vendors will develop packet-oriented gear with some optics integrated into it."
Perhaps unsurprisingly, Juniper says the latter approach will ultimately save telcos more money in the long run. A cost analysis from Juniper reckons that that a packet-centric integration solution (i.e. MPLS-based with OTN switching) would cost 65% less than a traditional optical network, while an optical-centric solution (i.e. hybrid router and MPLS/OTN switching) would save just under 50%.
That said, Ceuppens admits the cost model makes specific assumptions that won't apply uniformly to different networks.
Which is as well, since operator decisions on a packet-optical convergence strategy will be determined by the architecture already in place, says Mikdashi of ECI.
"Some operators want a very intelligent Ethernet network with some basic WDM capabilities, but others that already have a complicated optical layer want an intelligent optical network with mesh capabilities, ROADM, etc, and a basic Ethernet layer," he says.
It also depends on what TDM and Ethernet services they have at the time, Mikdashi adds. "If their network is dominated by Ethernet and IP services, they can deploy a packet-optical network that's more oriented on Ethernet and is stronger on Ethernet capabilities than optical. So there's no one rule of thumb."
Ceuppens agrees, and adds that despite the fact that operator interest in packet-optical convergence is high, actual implementation is going to take time as telcos weigh their options on when and where to flatten the packet and optical layers with minimal disruption to existing services.
"It will be slow because of the amount of legacy equipment and architecture carrying live traffic," he says. "As a service provider you don't want to mess with your customer. So very often they work with overlays - create that NGN and then migrate customers over to that. It's only when it's proven that the new network is cost-effective that they'll make the effort to migrate customers over to it."
There will also be one other convergence issue that telcos will have to resolve, adds Mikdashi - interdepartmental convergence.
"For many Tier 1 operators, the packet divisions and optics divisions are usually separate," he says. "So telcos do have to look at how this is going to affect those two divisions."
Changaroth of Nortel agrees. "Organizationally SDH/optical and IP operate as separate business units. So operators will be more challenged to drive them to work more closely together. Some are already doing it, but it will still be an issue for many."