Fiber's popularity grows for Fibre Channel applications
The acceptance of storage area networks (SANs) incorporating expanded link lengths has begun to put the fiber back into Fibre Channel, says a manager at one component manufacturer. New transceiver and Gigabit interface-converter (GBIC) technology will evolve to keep pace, particularly as applications shift to 2 Gbits/sec and standards bodies begin to discuss a move toward 10 Gbits/sec.
"Over the last couple of years, we've seen a lot more inclusion of optical in the systems. At 1 Gbit/sec, more than half of the market right now is using optics," says Brian Gregory, product-marketing manager for optoelectronics at Molex Inc. (Lisle, IL). Molex makes components for both copper- and fiber-based Fibre Channel applications. "All the new designs, as far as I can tell, are incorporating optics. There are very, very few people out there that don't have a lot of optics on their systems and aren't making a lot of plans to incorporate even more."
Gregory sees growth in the demand for equipment with optical interfaces as the length of SAN connections increases. SANs enable network managers to link individual data-storage devices into shared networks. The ability to create such networks over long distances and at potentially multigigabit speeds has enabled Fibre Channel to replace SCSI as the interface of choice for storage applications.
The use of fiber optics in Fibre Channel, however, has lagged that of copper, which has been the traditional medium for providing storage devices with communications capability. With network planners now looking to extend the distances they expect SANs to cover, fiber is becoming increasingly popular, says Gregory.
"There are a lot of situations where fiber makes a lot of sense," Gregory explains. "If you have very large storage devices, especially with some of the new Internet stuff that's coming on where people have massive storage needs, it lets you space those around and move them to different areas. Especially if you need to back up stuff--if you have a second location where you're taking your data and backing it up, it gives you a lot more options. For 10-km singlemode [applications], people can use dark fiber." For example, he notes that banks in New York are leasing dark fiber to connect their main computers in the city with backup storage facilities in New Jersey. Such applications will drive greater demand for optical technology in Fibre Channel systems, particularly as speeds increase.
"We see a lot of movement toward the 2-Gbit/sec Fibre Channel. Pretty much every account you go in right now is discussing 2-Gbit/sec Fibre Channel in one way or another. I think that will have a larger optics emphasis, but there will still be a copper version of it, I'm sure. So we think you'll continue to see a transition," Gregory says. He predicts that future versions of Fibre Channel will drive this transition even further. "I think it's likely that the next rate move will go from 2 gig to 10 gig--and 10-gig Fibre Channel will be heavily optics. I don't know if they'll do an electrical version at all at that speed," he says.
The typical economic differences between copper and fiber--with the former generally costing less money to deploy--hold true in the Fibre Channel environment. But economics frequently are not the determining factor in the decision between copper and fiber. "If you have GBIC ports on a Fibre Channel system, and you only need to go 5 feet, there are lots of people who do that optically, and I'm not sure what drives that. But if you put in a copper GBIC and a short cable, it does cost less," Gregory says. "But then again, if you need to go a long distance, it becomes exponentially more difficult each additional meter you go out. So it really depends on transmit distances."
As systems vendors seek to respond to their customers' requirements, the components manufacturers have begun work to decrease the cost of optical technology. This can be seen most clearly in transceiver development. "We also see, with people introducing Fibre Channel switches now, their port densities are going up, and there's a big drive toward small form factors," he reports. "We're going to be supporting both the LC and the MT-RJ, and we're seeing the market split between those two form factors. Up until about four months ago, the MT-RJ really seemed to have the market, and I've seen a huge swing toward LC [since then]."
The popularity of the LC may be due to both a pull from the systems side and a push from the component side, as design complexities inherent with manufacturing a transceiver around the MT-RJ have given some manufacturers problems. "If you look at the center-to-center spacing between the left and right fiber in an MT-RJ transceiver, it's very close together," Gregory explains. "The laser and photodiode that are farther back in the transceiver are widely spaced, and you have to have some sort of light pipe or guide that brings it down to those close center-to-center spacings. And that's a tricky design and it makes production more difficult. If you look at all the optical-transceiver vendors on the market, there are, I think, quite a few more that are supporting the LC now, just because it's a straight shot. You put in the laser and you can guide it straight out in a manner that's very, very similar to the SC."
These design issues make it difficult to predict the economics of transceivers using the two designs. "If you want a low-cost optical transceiver right now, it's certainly a debate as to whether the MT-RJ or the LC is lower cost for the transceiver manufacturer," Gregory explains. "Because there are some design issues that make the MT-RJ more difficult--but then again, there are some things long-term that may allow tooling to bring down the cost and bring it into parity or maybe even slightly below. And it's really difficult to determine how that will fall out."
Other design factors will also affect component costs, particularly as system vendors attempt to meet the needs of customers moving to 2-Gbit/sec Fibre Channel applications. "A difficulty that really comes into play right now is that for Gigabit Ethernet and Fibre Channel parts at 1 gig, people usually make the same device and then do a screen--final test will be for Fibre Channel or the final test will be for Gigabit Ethernet," Gregory says. However, Gigabit Ethernet specifications call for a bandwidth limit on the receiver to sharpen performance at the 1-Gbit/sec speed. "If you make a single part for Fibre Channel that accommodates both 1 and 2 gig, that part probably will not be a good Ethernet part because of the bandwidth limit on the receivers," Gregory reports. "So that makes the designs a little more difficult, and it means that some of the products are actually going to split and be really different."
Component vendors are studying the problem now. "There are some ways that it's possible to get around that problem as well. That may not be an absolute brick wall, but it's something that everyone is having to look at carefully," Gregory says.