MSAs, 10G, and beyond
There is now a burgeoning number of module multisource agreements (MSAs), each purporting to offer smaller form factors, lower power consumption, and enhanced ease of use. At the present time, the MSA of the future seems to be XFP, but a new form factor called SFP+ is just emerging to challenge it. Some of the major switch vendors are already adopting SFP+ for their next-generation boxes. At the same time, the IEEE group that deals with 10-Gbit Ethernet (10GbE) is coming up with new options that can reduce costs through improved transmitter design and the enabling of lower-cost copper and multimode fiber media. The two most exciting options coming out of the IEEE are the 10GBase-LRM standard for multimode and the 10GBase-T standard for copper, which purport to bring down the cost of carrying 10GbE signals over reaches that account for 80% to 90% of the addressable market for 10GbE.
These MSAs and IEEE 10GbE standards are separate market influences, but interact in important ways. While the IEEE standards define the optics and media, MSAs are more concerned with form-factor issues. The IEEE standards are specifically standards for 10GbE, while the MSAs also apply to OC-192 and 10-Gbit/sec WDM. However, although the newer MSAs could be used for other protocols, much of the discussion of MSAs is in the context of Ethernet, simply because the volumes of 10GbE modules likely to be sold are so large when compared with any other protocols.
The growth in Ethernet port sales is fueled by the need for aggregating the surging number of GbE ports on both business and even consumer computers. GbE has now taken over from “Fast” (i.e., 100-Mbit/sec) Ethernet as the garden variety of Ethernet, and network managers at large enterprises have realized for a few years that their backbone LANs cannot satisfy these hungry beasts. As our forecasts later in this report set out, annual sales of 10GbE transponders are expected to reach $2.3 billion worldwide by 2011.
Long-term opportunities for 10GbE will ultimately include the gradual replacement of GbE as the mainstream LAN variety; with the increasing use of high-speed digital video in the home, it would be no surprise to see 10GbE enter the residential marketplace as well. However, there are more immediate opportunities in 10GbE, and these consist in providing the best 10GbE solution for the 90% of the market that needs a reach of 300 m or less.
The conventional wisdom is that this part of the market is largely going to be served in the future by the IEEE LRM standard in conjunction with the XFP module, with the shortest reach requirements being met with 10GBase-T. And it is a reasonably safe bet for firms to follow this conventional wisdom. However, one risk with this approach is that it will succeed all too well and that a firm that follows the road to XFP, LRM, and T will soon find itself in a commodity business. This may make little difference to established suppliers that manufacture in high volumes, but it is clearly not the path to success for smaller firms whose competitive advantages must lie in innovations. On the copper side of the house, Cat 6/6a and Cat 7 cabling needed for 10GBase-T is very expensive. Both these challenges may leave the market open to other ways of doing things, such as plastic optical fiber. Finally, XFP may face a serious challenge from SFP+ somewhere down the line, if only at the lower end of the market.
Meanwhile, back in the public network, much has been made of the “death of SONET.” This does not yet seem to be as imminent as some people feared or hoped. Expect that the future OC-192 story will be more one of continued steady-but not dramatic-growth as telcos upgrade their networks. Dazzled by predictions of astronomical traffic and revenue growth, the service providers overspent during boom times and won’t be fooled again, even though they are now seriously looking to upgrade their facilities and install next-generation networks. Expenditures are not likely to be lavish, and OC-192 will only be bought when the existing connections are straining. If OC-48 will do the job then OC-48 is what they will buy, and sales of OC-48 ports-which are really just a commodity at this point in time-have proved fairly vibrant in recent years. It could be argued that this is somewhat shortsighted, but the accounts departments will not agree to anything that looks like excessive capital expenditure.
All this might suggest that OC-192 is not really in need of a successor and raises issues about whether OC-768 is worth pursuing. As the optical networking business went into decline, OC-768 more or less symbolized the irrational exuberance of the period, and many of the firms that were making OC-768 technology their core competency dropped out. Now it seems 40G is back, although most of it is either being used in network trials or R&D. The one “commercial” application for OC-768 seems to be for router-to-add/drop multiplexer connectivity; management of one big IP pipe is easier, which makes up for the lower cost of deploying four OC-192s instead.
Although OC-768s deployed in this way are few and far between at the present time, OC-768 on routers suggests strongly that 40G will present an opportunity in the future. This is because it is exactly in this application that OC-192 got going. In addition, since SONET/SDH is not going away any time soon, it seems to us likely that carriers will begin to upgrade their existing networks using OC-768 in a piecemeal fashion. In some cases these carriers will skip the OC-192 stage as OC-768 becomes more affordable. The recent announcement of an OC-768 backbone by AT&T suggests that the deployment of OC-768 on a strategic basis may also be a possibility.
One advantage of offering 40-Gbit/sec products is that it gives a leg up for entry into the next generation of networks that will operate at 80, 100, or 160 Gbits/sec. Many of the technological problems with which manufacturers of OC-768 products must cope will be the same ones that will be faced in commercializing these faster networks. It may turn out that approaches for OC-768 are scalable to even faster networks.
The table shows CIR’s latest forecasts of 10- and 40-Gbit/sec modules.
Lawrence Gasman is president of Communications Industry Researchers Inc. (www.cir-inc.com). The information for this report comes from CIR’s new study, “The Market for 10G & 40G Modules.”