PMC readies OTN linecard ICs
As OTN spreads from the backbone to the access network, service providers and enterprises look to a new generation of line-card silicon.
By Ron Wilson, Executive Editor -- EDN, July 21, 2009
The entire communications industry, from investors to carriers to service providers and enterprise IT managers, has been discussing the shifting traffic loads on the global communications network. There is no debate that we are handling an exponentially growing load of data, image, and even HD video traffic over a network that was intended for digitized voice.
The owners of the backbone are responding, moving rapidly from Sonet/SDH to the OTN (Optical Transport Network), ripping out old equipment and dropping new OTPs (Optical Transport Platforms) and ROADMs (Reconfigurable Optical Add/Drop Multiplexers) into their existing fiber networks and adding fiber capacity.
The shift, according to PMC Sierra Director of Marketing Babak Samimi, is a simple matter of efficiency. Sonet/SDH was conceived and optimized to handle time-multiplexed voice traffic concentrated onto T1/E1 lines. As data flows became more intense, the network responded with virtual concatenation to provide the appearance of larger channels, but it did so at the expense of complexity and inefficiency. OTN, in contrast, was intended to be a friendly environment for packetized traffic, and especially IP (Internet Protocol) traffic. "You get much more efficient channelization of traffic, and in particular much better mapping of non-deterministic data types," Samimi observed. The increased efficiency has become mandatory as the traffic has grown and shifted toward IP, driven by such user-level trends as adoption of Cisco's virtual networking concept and the data-consuming iPhone.
Samimi continued to state that the impact of this change is already starting to spread beyond giant carriers and their all-optical backbones. End-to-end carriers like Verizon or China Telecom are pushing OTN out into their access networks. And there are first indications that independent service providers such as the CLECs (Competitive Local Exchange Carriers) and even enterprises will join in, translating their traffic from its native mode to OTN as near the source as possible.
"We are seeing the beginning of a total end-to-end OTN world," Samimi suggested. "We will see all kinds of traffic, from existing voice and data to video and even extended storage-area networks [SANs], mapped onto the OTN. This has begun, but it will be a ten-year transition, not an overnight switch."
This new stage of the transition creates a hole in the silicon infrastructure for OTN. Chips so far have primarily been designed to live at the edge of the backbone itself, transporting already-concentrated data flows into the OTN space. With the spread of the network, the industry needs silicon that can take in the variety of legacy traffics from their sources—Sonet/SDH streams, video streams, traffic from SANs, Ethernet traffic, as well as OC-192, 10 Gbit Ethernet, FibreChannel, and the like—and concentrate them onto an OTN connection.
But wait, there's more. Samimi pointed to his remark about a 10-year transition. Many of these concentrating devices—be they OTPs on the network perimeter, connections on legacy Sonet rings, DSL central offices, or the new micro-OTPs, micro-MSPPs (Multi-Service Provisioning Platforms), or transponders—will begin life in a primarily legacy world and see the environment around them gradually migrate to all-OTN.
Ideally in such a scenario, you would want silicon that can support a wide array of new and legacy clients and attach to all of the major high-capacity nets: OTN, SONET/SDH, and fast Ethernet. That is what PMC Sierra is offering in the HyPHY chip family.
The two initial offerings are the HyPHY 10G and 20G—roughly designated by throughput. The chips are intended for any-rate, any-service line cards to bring the existing world into the OTN network. The line cards can serve in micro-OTN boxes, OTPs, or ROADMs. In concept, you can stay with one line card even as the incoming clients and the outbound network change, because all the service combinations are supported in the programmable chip.
This requires a fairly complete set of functional blocks. The HyPHY 20g, for instance, includes a pair of XFP SerDes, a pair of SFI 4.1 SerDes sets, and 16 SFP SerDes on one side, and Packetized OTN, SONET/SDH, and Packet-Ethernet/IP SerDes banks on the other side. In between, there are the necessary framers, muxes, switches, MACs, and OTN and SONET support blocks to get everything to talk to everything. There is a controlling CPU core, and interfaces to external memory and MCU devices.
Functionality of the interior blocks has been tuned to the transitional environment Samimi describes. For instance, the chip can perform concurrent OTN and SONET/SDH grooming for environments in which it must serve both networks. The SerDes are designed to be rate-agile without gaps, in order to support the atypical rates the chip may encounter in video networks or SANs. And the hardware supports Timing-over-Packet and Ethernet OAM (Operation, Administration, and Maintenance) to provide the time-referencing and services necessary for wireless backhaul.
PMC believes the HyPHY chips are currently unique and will give the company a head start as carriers, service-providers, and enterprises adapt to the OTN world. Starting at $550 in less than 5K quantities for the smaller HyPHY 10g chip, the devices will not be inexpensive as chips go. But they should lead to very substantial cost-of-ownership savings at the line-card level.
