BRIEFING

SONET vs. IP over Photons: Debate and Reality

from the March 1999 issue of Business Communications Review, pp. 14–16

Reasonable people are disagreeing—sometimes vehemently—about the relative merits of implementing IP directly over photons, instead of continuing to enhance the entrenched SONET base. Proponents of IP-over-photons, such as Graeme Fraser, VP and GM of optical internetworking at Cisco Systems (www.cisco.com), argue that their approach best matches the requirements of the Internet, where chunks of data can move rapidly by cell or packet switch. SONET, they point out, was made for circuit-switched, voice-oriented networks; as data volumes outdistance voice, so should IP over photons be replacing SONET. (For a different view of circuit and packet backbones, see Dave Passmore's column in this issue, pp. 19–20.)

"Ultimately, it comes down to a decision the carrier has to make as to whether it wants to participate in the growing Internet or not," said Fraser. Of course, it's worth noting that Cisco has not made SONET equipment but does make routers that would be instrumental in IP-over-photons networks.

On the other side of this argument are SONET defenders such as Bill Miller, VP of broadband services at Fujitsu Business Communications Systems. Miller maintains that a move to IP over photons will result in carriers sacrificing protection and reliability. He compares the situation to the reliability of a landline phone as opposed to cellular, and predicts that the builders of IP-over-photons networks won't be willing to pay for the 99.999 percent reliability SONET offers.

Again, there's corporate self-interest at work here: Fujitsu played an instrumental role in developing the original SONET standards and has a major market in SONET equipment. Such recent SONET innovations as two- and four-fiber bidirectional line-switched ring (BLSR), which Fujitsu is promoting, would not necessarily be used with IP over photons.

So Who Needs SONET?

In making their case, proponents of IP over photons argue that the redundancy that guarantees SONET's reliability represents overkill that keeps the network from using a large portion of its resources. "Half of the bandwidth in a SONET ring is set aside doing nothing," said Fraser. "In a data network, we can take advantage of parallel paths and use the 'meshiness' of the network to reroute traffic around the problem without dedicating half the bandwidth to it."

As a result, the cost of delivering IP over photons can be an order of magnitude lower than SONET, Fraser contended. "IP over DWDM can eliminate several layers over SONET and ATM, it can eliminate a lot of that equipment, it is a very efficient backbone," he said.

What's more, carriers that had committed to SONET are beginning to see these economics for IP over photons; recent examples of major carriers using the technology include AT&T, Sprint, Enron, Frontier, Canarie (a Canadian network) and Global Center. "Until recently, I would have said that our main stumbling block to acceptance was skepticism about how well the technology would actually work in the field," said Fraser. "That is now largely gone."

This doesn't mean that IP over photons won't have to prove itself. "The real test is whether we can create an end-to-end optical Internet operating from OC-3 to OC-48 and build systems around an optical Internet backbone," Fraser added. "Our focus over the next year will be to continue to revolutionize the on-ramps," to make the network IP-centric and IP-optimized.

In Defense of SONET

Fujitsu's Bill Miller takes issue with the argument that SONET costs more than IP over photons, asserting that the costs are roughly equal. The enormous routers it will take to replace circuit switches will, Miller argues, wind up being just as expensive as those switches. Also, he contrasts SONET equipment's long depreciation cycle with the continual need for new equipment on the part of IP-over-photon networks.

SONET can continue to handle new challenges, Miller said, pointing to virtual private networking capability that is now being provided on SONET. The technology's hierarchical scale also has led to a smooth ramp up the data rate ladder—now to OC-192, with potential for OC-768. SONET's durability leads Miller to describe it as "the John Glenn of technologies." That's important, because the public network infrastructure isn't going to be swapped out anytime soon, as one of Miller's colleagues observed: "Class 5 switches are going to be here for another 20 years," said Greg Wortman, senior director of marketing at Fujitsu Network Communications.

Miller contrasted the current public network infrastructure with IP over photons, which he said does not have the support personnel to really help carriers. Fraser acknowledged that this is a problem: "We have to get enough people literate," from design engineer to technician, he conceded. Yet proponents like Fraser also believe that the area is exciting enough, and the potential great enough, that bright new minds will be attracted to the field and that these problems eventually will be overcome.

Truth Lies in the Middle

The reality of the situation is somewhere between these two points of view. In fact, in a bit of a departure, Cisco announced that it is investing in a new SONET manufacturing concern, Cerent Corp. Cerent gear is packet-based and can accommodate any network service over any optical bit rate, the company says.

Moreover, conversations with leading members of the Optical Internetworking Forum (OIF) make it clear that IP over photons will not exist without a SONET frame or interface, at least in the near term. What is being replaced near-term are SONET multiplexers, which have long been derided by fiber purists anyway, because they reconvert the signals back to electronics before boosting them photonically.

Therefore, the lack-of-reliability argument is a red herring, at least for now. Andrew Greenfield, OIF's president, noted that all IP-over-photons systems now operational are SONET framed. He also cautioned that IP-over-photon deployment is left pretty much up to the vendor and carrier developing the network, leading to proprietary implementations.

In general, there is a need to better understand what is happening, said Greenfield. "For example, where exactly does the data reside in bit streams? We definitely need to work on mapping packets to bits," he observed.

These uncertainties undermine some of the arguments in favor of IP over photons. Because there is no network model that can be set down side by side against a model SONET network, the assertion about IP-over-photons' order-of-magnitude cost savings is impossible to verify. Indeed, the fact that IP over photons is a new technology and equipment has not matured pricewise would argue against that type of savings.

What does ring true is the inevitability of IP over photons. Even Miller acknowledged that the transition will happen—it is just a matter of time. Fraser characterized the resistance to IP over photons as being "normal," noting that, "People have built their careers around [SONET]." He further recognized that it is going to take varying amounts of time for carriers to adopt IP over photons, "depending on who you are and what environment you are in." For carriers attempting to differentiate themselves, fuller implementation may come earlier, while for others it will be a slower process.

Top of Page