by Stephen M. Hardy
I was talking to a spokesman for the OIF the other day about a new effort to create a board-level interface that would enable future optical transceivers to hand 100-Gbit/sec traffic to SerDes devices and other chips. The OIF decided to tackle the challenge because the other major standards bodies working on 100 Gbits/sec-the IEEE and ITU-T-weren’t likely to consider the subject in their deliberations. The spokesman conceded that work on this follow-on to the SFI-4 (currently called the SFI-X, although the spokesperson predicted that a number would replace the “X” in the interface’s name at some point) would likely progress only as quickly as did the IEEE High Speed Study Group’s and the ITU-T’s efforts at 100-Gbit/sec standards, since it’s tough to know what kind of signaling parameters you need to deal with exiting an optical transceiver to the board if you don’t know what the signals look like coming in from the fiber.
So I asked him whether, to make such an interface more immediately useful while 100-Gbit/sec transmission parameters remained a matter of debate, a variant of the SFI-X targeted at 40-Gbit/sec optical modules might make sense. The spokesman ruefully noted that the SFI-X developers would likely take their cues from the industry at large, and right now the industry seems uncertain whether, with the advent of 100-Gbit/sec standards work, 40 Gbits/sec was a technology whose time had passed even before it had arrived.
As readers of my article on the next generation of 40-Gbit/sec technology will discover (see “Vendors Align Technologies to 40G Strategies” on page 30), suppliers in the 40G space are fully aware that certain sectors of the optical communications community-namely, those with datacom backgrounds-view their offerings as irrelevant in the grand scheme of things. Network architectures and traffic patterns continue to swing towards packet-based Ethernet and IP. The old, “telco” way of doing things is now obsolete, replaced by a new paradigm that comes from the datacom world, say some. That means network upgrade schedules will run on the Ethernet clock, where bandwidths rise by a factor of 10 each generation, rather than four. So from 10 Gbits/sec, 100 Gbits/sec represents the next logical step, not the “archaic” 40 Gbits/sec of the SONET/SDH regime.
Not surprisingly, vendors of 40-Gbit/sec technology beg to differ. And you know what? I’m with them. No one doubts that 100-Gbit/sec networking will arrive and thrive. However, it seems clear to me that there is a demand now and even in a 100-Gbit/sec world for 40-Gbit/sec technology.
Let’s start with current requirements. It’s true that increases in packet-based traffic drive higher speeds; the 40G vendors concede that the demand for their technology currently rests on the 40-Gbit/sec interfaces companies such as Cisco have installed on their large packet routers. You can certainly foresee that Cisco and other router vendors will add 100-Gbit/sec interfaces to their boxes when they become available, thus sparking a demand for 100G networking. But we aren’t going to see 100-Gbit/sec standards any time soon-and even when we do, there’s no guarantee that the first generation of technology based on those standards will approach the price points carriers will demand.
Meanwhile, those 40-Gbit/sec interfaces are available today-and carriers have begun to use them. As one of my sources pointed out, carriers don’t increase transmission rates because they like installing cool new technology; the cool new technology always costs more than what they’re currently using. Carriers will move to 40-Gbit/sec transmission for the same reason they upgraded from 2.5 to 10 Gbits/sec: Bandwidth demands won’t give them any choice, including the choice of waiting around for economical 100-Gbit/sec transponders to appear.
And even when such 100G transponders do reach the field, 40G technology will still have a niche. Despite the disparaging tone with which they sometimes use the term, 40-Gbits/sec definitely makes sense from a “telco” perspective. (Who outside of the data center/private network space do they think is going to deploy high-speed transmission technology anyway-ISPs?) For each high-speed lane paved with the latest generation of dispersion-tolerant fiber there are several that currently struggle to support 10-Gbit/sec wavelengths. Many of these links will reach capacity exhaust just as surely as will the gleaming new major backbones, and 40-Gbit/sec technology (provided the price points have come down sufficiently) will provide the most economical and perhaps the only technologically feasible alternative to laying new fiber.
So to the OIF and other standards organizations I say go ahead and plan for 40-Gbit/sec technology. You won’t be wasting your time.