Fibre Channel promises jolt of technological adrenaline
Users can choose from a variety of performance levels
Fibre Channel Association
Fibre Channel is a fast interface that can deliver gigabit performance over either copper or fiber-optic media. Interface speeds that range from 133 megabits per second to 1.062 gigabits per second have already been defined, and with the development of the new document defining Fibre Channel extended physical, those data rates will increase to 2.134 and 4.268 Gbits/ sec.
Behind these numbers is a hybrid technology. Fibre Channel takes aspects of input/output channels and networks to create a very high performance interface. Like a channel, it is hardware-intensive and bullet-proof, works at high data rates, and offloads most station-management responsibilities. But like a network, it provides excellent service in environments of indeterminate connections and works over long distances. Central to Fibre Channel`s high performance is that it provides memory-to-memory data transfers through direct connection to a nonblocking switched communications fabric. Fibre Channel can achieve gigabit-per-second data rates because all channel and network protocols use buffers to hold the data they send and receive. Fibre Channel simply takes the contents of the transmitter`s buffer and blasts them directly to the receiver`s buffer. What kind of data happens to be in the buffer--commands, packets or frames--is immaterial. The transfer, itself, burns no central-processing-unit cycles, requires no decoding and adds almost no overhead.
Fibre Channel`s architecture is completely collapsed into hardware, which yields tremendous benefits in heterogeneous environments. Standard network types, such as Ethernet, token ring or fiber distributed data interface, require major retrofitting to interoperate because they all have frames that are dependent on the topology of the network. Fibre Channel maintains the size and structure of its frames across all media and topology types so that these physical dependencies and their consequent incompatibilities disappear.
In addition to high and scalable data rates and topology independence, Fibre Channel promises a number of benefits to users and developers looking for the next jolt of technological adrenaline.
One of the greatest benefits is protocol independence. Because protocols are invoked either before data enters the sender`s buffer or after it enters the receiver`s buffer, they have no effect on the transfer, itself. Fibre Channel communication is stackless and flexible. Users and developers can choose from a wide range of performance levels. They can also select different types of switching (packet or circuit), connection (point-to-point link, arbitrated loop, full switching fabric) and distances between nodes (inches to 10 kilometers). Users can also choose to multiplex different channel and network connections, line speeds and communications protocols over a single port.
The result is lower costs. With such a powerful and flexible interface, system designers need fewer interfaces to satisfy users` requirements. With station management offloaded to the switching fabric, the interface can be simpler. With Fibre Channel`s ability to connect directly to that fabric, the need for and expense of an intervening central processing unit are obviated. And with its small footprint and ability to do a lot over a single port, Fibre Channel has an advantage in an environment where input/output slots account for the majority of the cost and market differentiation in a typical system.
A voice of authority
In the early days, these advantages were at least partially compromised by ill-informed coverage in the trade press that said Fibre Channel was exclusively a local area network technology and that Fibre Channel was a direct, zero-sum competitor with asynchronous transfer mode. In an attempt to correct these misperceptions, the Fibre Channel Association was formed in August 1993.
"There was no voice of authority to speak for Fibre Channel, so we decided we`d better create one," says Jeff Silva, program manager for Fibre Channel market development at IBM`s RISC System/6000 division in Austin, TX, and the chairman at the Fibre Channel Association. Informal confabulations among five engineers attending the regular American National Standards Institute X3T9 Committee meetings indicate the association has grown steadily--from 19 members in 1993 to 57 members, including Apple, Hewlett-Packard Co., IBM, Intel and Motorola, and a number of noncommercial research institutions such as the Argonne and Lawrence Livermore national laboratories and the Johns Hopkins Applied Research Laboratory.
These members did not come together to develop the standard; that is ANSI`s job. FCA`s mission is to raise awareness of Fibre Channel technology within the industry and among consumers. "There are a lot of vendors supplying these products," says Silva. "It`s in all of our interests to have an educated market."
The state of the standard
The two major questions regarding Fibre Channel are those asked of every technology that purports to be a standard: "What is the status of the standard document?" and "Will products from different vendors interoperate?"
From the late 1980s, when the idea for the standard first took shape, until 1994, Fibre Channel was governed by ANSI`s X3T9 Committee. In 1994, however, the group was renamed the X3T11 Committee.
The X3T11 Committee has presided over a unanimous vote in favor of approving the standard known as FC-PH, which defines all of the nuts and bolts of the physical layer, but it has not yet published the document describing it. Roger Cummings, senior advisory engineer at Storagetek, Louisville, CO, and chairman of the X3T11 Committee, says this is a formality. "X3T11 was unanimous," he says. "Everybody has made it clear they want this to be a standard."
When asked if there were any disagreements that could threaten Fibre Channel`s ultimate adoption, Cummings said there were not, but added, "The thing we hear most often is that it`s very complex. This is a justified criticism, given that we`re trying to be appropriate for everything from 3.5-inch disk drives to supercomputer I/O channels and network attachments."
The ultimate test
The complexity of the standard is well known and has engendered an ongoing effort to develop interoperability profiles. Interoperability is the ultimate test of any standard, and the development of profiles to make it possible is the subject of intense effort in the Fibre Channel community. At present, the profiles are developed by the Fibre Channel Systems Initiative, a group dedicated mainly to this task and wholly separate from FCA. By the end of 1995, FCA will take over the responsibility of developing and maintaining these profiles. These profiles are designed so manufacturers can ensure interoperability, without having to either implement the entire standard or guess which of its elements are essential.
"Fibre Channel has an incredible number of options," says Ed Frymoyer, the Fibre Channel Systems Initiative program manager at Hewlett-Packard`s Information Networks Division in Cupertino, CA. "Unless you have a profile of some kind, it`s a daunting task to make Fibre Channel-compliant interoperable hardware. And implementing the entire standard would make products more expensive than what people are willing to pay." The FCSI`s profiles provide a straightforward road map that describes which parts of the standard must be implemented to ensure interoperability among systems designed for specific applications.
So far, they have published three profiles: the common feature set, which includes the baseline functionality that all profiles must have to be compliant with the Fibre Channel standard; small computer systems interface over Fibre Channel; and transmission control protocol/interconnect protocol over Fibre Channel. In the near future, FCSI will publish one more profile for short-distance clustering. Four other profiles are being worked on by other industry groups: interoperability with fabrics; arbitrated loop communication; connection with Fibre Channel arbitrated loop and direct disk-attach.
Some standards, such as the fratridial conflict over Fast Ethernet, are not immediately accepted, but there is wide acceptance of Fibre Channel. "This is a standard people are flocking to in droves," says Frymoyer. According to Silva, "The fundamental characteristics defined in the Fibre Channel technology make it so flexible that it can be accepted by major vendors as a bona fide standard in each of three market areas: as a channel, a disk I/O and a network backbone." Silva adds that we are already starting to see companies in these markets embrace Fibre Channel, even in its present undocumented state. Emulex, one of the growing players in Fibre Channel, just published an annual report that bases their strategy in the LAN and I/O arena on Fibre Channel. In November, Hewlett-Packard, Quantum and Seagate announced support of Fibre Channel`s arbitrated loop as a disk-drive standard; in addition, Seagate recently introduced a product that implements Fibre Channel as a disk drive interface. These two recent developments place Fibre Channel "squarely in the middle of the disk drive game," Silva continues.
"You can see the eagerness of the industry to jump on Fibre Channel from the channel perspective, from the network perspective and from the disk drive perspective," he says. He notes that Ancor Communications and Radway International Ltd. have announced bridge/routers that extend the connectivity of the Fibre Channel switched infrastructure out to legacy LAN environments, establishing the standard as a tool in the ongoing effort to upgrade the performance of existing LANs.
Between 85 and 100 companies manufacture Fibre Channel products, in cluding multiple vendors of chips, disk subsystems, bridge/ routers, network switches and bus adapter boards. "We`re starting to see an interesting explosion of Fibre Channel in several different directions," Silva says. "A lot of companies are saying the Fibre Channel is going to take them into the next century."
The final question about Fibre Channel is: "Whose interests will be hurt by a large-scale adoption of this technology? Slower channel technology`s with larger footprints are likely to be adversely affected.
"Many people will think of us as a natural successor to small computer systems interface," says Silva. "It is not going to go away, but we will probably see SCSI, along with enterprise systems connection and high-performance parallel interface, fade over time."
Slower networking environments are also at risk, but here the situation is complicated by the maelstrom of development in this arena and the resultant information entropy. "As a backbone technology, yes; FDDI will possibly be affected. But it`s my personal belief that it will be a long time before we see Fibre Channel to the desktop. There`s just too much confusion."
Two forces are likely to increase the acceptance of Fibre Channel. One is the drive toward interoperability. The effort to develop profiles is ongoing. Hew lett-Packard, IBM and Sun have already demonstrated interoperability among their machines in the real world, using a live video signal. And FCA has announced that the Lawrence Livermore National Laboratory has agreed to serve as a disinterested test site. It will develop programs and suites that test for standards compliance per se, and for the degree of interoperability among various vendors` equipment.
The greater force is the insatiable demand for more channel and network bandwidth. Fibre Channel is a flexible and expandable technology that can accommodate growing demands. As long as applications` requirements for bandwidth continue to expand at the present exponential rate, Fibre Channel will have a future. And with the succession of new technologies that will foster that demand--among them the widespread adoption of distributed applications based on enormous multimedia files containing voice, video and computer animation--that future looks secure. u
David McDougal is a technical communications writer for David McDougal & Associates, Redwood City, CA.