FIBRE CHANNEL TECHNOLOGY Data Transfer Solution for the 21st Century White Paper From Unisys Open Storage Solutions
FIBRE CHANNEL
TECHNOLOGY
Data Transfer Solutionfor the 21st Century
White Paper
From Unisys Open Storage Solutions
C O N T E N T S
1 Executive Summary
1 What Is Fibre Channel?
1 Does Fibre Channel Comply with ANSI Standards?
2 Is Fibre Channel Becoming Industry Standard?
2 What Can Fibre Channel Do?
4 How Does Fibre Channel Connect Network Nodes?
7 How Is Fibre Channel Organized?
9 What Media Do I Use for Physical Connections?
10 Is Fibre Channel Being Used Successfully Today?
12 What’s the Last Word on Fibre Channel?
13 Appendix A: Fibre Channel Quick Reference Card
F I G U R E S
1 Point-to-Point
2 Switched Fabric
3 Loop
4 Fibre Channel’s Five Layers
5 The SAN Solution
T A B L E S
1 Attributes of Copper and Optical Fiber Cable
A1 Fibre Channel and Related Standards and Technical Reports
A2 FCA and FCSI Profiles and Related Documents
A3 FC-PH
A4 DF_CTL Bits for Data Frames
A5 Header Fields by Frame Group
A6 CS_CTL Bits for All Frames
A7 Parameter Field for BSY and RJT
A8 “Device” Type Field Values of FC-4 Frames
A9 F_CTL Bits for All Frames (Word 2)
A10 Reject (P_RJT, F_RJT) Reason Codes
A11 Optical Feeds and Speeds
A12 Electrical Feeds and Speeds
A13 Class Characteristics
A14 Well-Known Address Identifiers
Executive Summary
Today’s businesses are demanding the ability to find and retrieve their data instantly, and
that data is much larger than ever before because of the tremendous growth in video, multi-
media, giant software suites, and other large files. What they need is Fibre Channel.
Fibre Channel is a high-speed data transfer technology. Its primary task is to transport
data extremely fast with the least possible delay. The Fibre Channel standard supports mul-
tiple file protocols and can use either copper or optical fiber cables. This f lexibility coupled
with the range of topologies available enable customers to custom-make the Fibre Channel
solution that best fits their needs.
Fibre Channel technology has brought to the market the high-performance, low-cost
communications pathway required by the new breed of data and communications-intensive
applications commonly used in networks. This technology also solves the distance, connec-
tivity, and performance problems seen in current systems using parallel SCSI. The high-
speed transfers can reach a rate of over 1 gigabaud per second—or 10 to 250 times faster
than today’s network protocols.
Fibre Channel is positioned for entry into the mass market and is ready to be the
number one system interconnection means of the future.
What Is Fibre Channel?
Fibre Channel is a high-speed data transfer technology—an integrated standards set devel-
oped by a committee operating under the American National Standards Institute (ANSI).
Fibre Channel’s primary task is to transport data extremely fast with the least possible
delay. Fibre is a generic term referring to all supported media types, while fiber refers to
the optical fiber transmission medium.
The Fibre Channel standard is proving the most useful in interconnecting servers,
storage devices, and workstation users. Its success is based on its transfer speeds, f lexible
topology, and f lexible upper-level protocols. Fibre Channel easily handles both networking
and peripheral input/output (I/O) communication over a single channel, resulting in fewer
I/O ports and fewer unique ports—the traditional bottleneck of other server connection
technologies.
Does Fibre Channel Comply with ANSIStandards?
In 1988, the American National Standards Institute X3T9.3 committee chartered the Fibre
Channel working group to develop a practical, inexpensive, yet expandable method of
achieving high-speed data transfer among workstations, mainframes, supercomputers,
desktop computers, storage devices, and display devices. The Fibre Channel working
group was assigned full ANSI committee status in 1994.
Fibre Channel Technology 1
Is Fibre Channel Becoming Industry Standard?
Fibre Channel is gaining industry support at a blinding pace as companies recognize the
technology as a solution to their data transfer problems. The industry now provides a full
system set of products in several markets. Early adopters in key industries, such as video
pre- and post-production, are enjoying significant economic benefits from their Fibre
Channel systems.
Technology giants such as Compaq, Hewlett Packard, Tektronix, SUN, Intel, Symbios
Logic, NEC, Unisys, Digital, SGI, Microsoft, and at least 100 others now have Fibre Channel
technology tied significantly to many of their products and market strategies. With Microsoft
support, Fibre Channel appears to be a winning strategy for Windows NT suppliers as well.
Fibre Channel has become the dominant standard for information exchange, beginning
with storage networking. In the future, as with other technologies that provided an industry
paradigm shift, Fibre Channel design improvements may well come from some of the
currently smaller participants.
Though no one can be sure how and where a technology will evolve, it is certainly clear
that Fibre Channel is here to stay and has already changed data transfer technology forever.
What Can Fibre Channel Do?
The main features of Fibre Channel are that it:
• Delivers speeds that are 2.5 to 250 times faster than existing communication and I/O
interfaces
• Overcomes today’s network performance limitations regarding error detection
and recovery
• Provides low-cost and reliable performance for distance connections
• Offers f lexible protocols and topologies to best leverage existing technology
investments
• Runs multiple protocols over a single network or loop
• Allows clustering of servers
• Enables multiple servers to share storage
• Scales from small numbers of peripherals attached short distances apart to large
numbers attached many kilometers apart
Delivers Fastest Transfer SpeedsData transfer needs in all types of business are growing rapidly. Files today combine data,
sound, graphics, and even video. The size of data is exponentially larger than in past years
and decades. Information technology (IT) departments the world over need to move this
information through their networks at lightning speeds.
Fibre Channel runs at four simplex speeds (actual data throughput): 100 megabytes
per second (MB/s) (1062.5 megabaud), 50 MB/s (531.25 megabaud), 25 MB/s (265.625
megabaud), and 12.5 MB/s (132.812 megabaud). Full-duplex transfers are allowed (READ
and WRITE simultaneously), which double the above throughput rates. Currently, though,
2 White Paper
few products support full-duplex transfers. For total bandwidth, a single 100 MB/s Fibre
Channel port equals five 20 MB/s SCSI ports. Fibre Channel provides a total network
bandwidth of about one gigabit per second.
These speeds are extremely helpful in solving the problems that growing file sizes
are creating. The high-speed transfers are available because of Fibre Channel’s f lexible
topologies and protocols, which are discussed at length in the “How Does Fibre Channel
Connect Network Nodes?” section of this paper.
Overcomes Network LimitationAt the same time that Fibre Channel transports data, it also performs error detection in
low-level hardware and microcode. Because Fibre Channel’s design is hardware intensive,
it offers a lower overhead for the host system than traditional networks. One reason for this
lower overhead is that LANs are quite software intensive to allow for station management.
Station management is how each station, or network node, recognizes error conditions on
the network, then provides the error management needed to recover.
Fibre Channel hosts do not have this burden. Rather, when congestion causes a data
transfer failure, Fibre Channel retries immediately without software intervention. Only the
more complex error recovery functions are passed back to the central system. Fibre Channel
relies on the contents of the frame header to trigger actions like routing arriving data to the
right buffer. Commands and responses are routed directly to and from command buffers
while data is routed directly to the memory allocated by the task that made the request.
Provides Distance Connection and Low CostThe Fibre Channel architecture provides a range of implementation possibilities extending
from minimum cost to maximum performance. The transmission medium is isolated from
the control protocol so that each implementation may use a technology best suited to the
environment of use.
Offers Flexible Topologies and ProtocolsFibre Channel ports can be connected in three distinct topologies: point-to-point links,
switched fabric configurations, or in a loop. Both optical and copper media are supported,
working from 133 megabaud per second up to 1.062 gigabaud per second. Also, using fiber
you can achieve distances of up to 10 km. Fibre Channel supports a great many protocols,
such as:
• High Performance Parallel Interface (HIPPI)
• Single Byte Command Code Set (SBCCS, also known as the Block MUX channel)
• Small Computer System Interface (SCSI)
• Internet Protocol (IP)
• Institute of Electrical and Electronic Engineers 802 (IEEE 802)
• Asynchronous Transfer Mode (ATM)
If these network protocols use Fibre Channel for their transport system, the original
higher-level command sets and functions are retained while the lower-level physical
interface is replaced by Fibre Channel.
Fibre Channel Technology 3
In addition to raw transmission speed, the use of a f lexible circuit/packet-switched
topology to connect devices is key to the high performance of Fibre Channel. Fibre Channel
is able to establish multiple simultaneous point-to-point connections. Devices attached to
a non-blocking switch do not have to contend for the transmission medium as they do in
some networks. Through its intrinsic f low control and acknowledgment capabilities, Fibre
Channel also supports connectionless traffic without suffering the congestion of the shared
transmission media used in traditional networks.
How Does Fibre Channel Connect Network Nodes?
The key to Fibre Channel performance and one of its most valuable features is its
f lexible topology, which offers three distinct interconnection choices:
Point-to-Point TopologyAs shown in Figure 1, the simplest of the three topologies is point-to-point. This method is
a single connection between two servers, or a server to storage. The point-to-point topology
uses a single, full-duplex cable between the two systems.
Point-to-point connections have several advantages. This topology has the highest
bandwidth efficiency possible because there are no devices interfering with the connection.
For the same reason, this topology has the lowest latency (delay) of all three connection
choices. Point-to-point connections are also the easiest and the least expensive way to use
the Fibre Channel technology.
The disadvantage of the point-to-point topology is that this is a very limited use of
Fibre Channel. Exactly two devices are connected—no more. In addition, both devices must
share the same data link speed and the same cabling scheme.
The best uses of point-to-point connections are when transferring extremely large
amounts of information specifically between two systems (either computer-to-computer
or computer-to-disk). When volume, cost, speed, and reliability are the mitigating factors
in data transfer, this topology may be the best choice.
Switched Fabric TopologyA switched fabric topology provides the greatest interconnection capability and largest total
aggregate throughput of the three topologies discussed. In this interconnection method,
each server or storage subsystem is connected point-to-point to a switch and receives a non-
blocking data path to any other server on the switch. This setup is equivalent to a dedicated
connection to every server. As the number of servers and storage subsystems increases to
occupy multiple switches, the switches are, in turn, connected together. Multiple connection
paths between switches are recommended to provide circuit redundancy and increase total
bandwidth. Figure 2 illustrates this connection method.
4 White Paper
Figure 1.Point-to-Point
An obvious advantage to a switched fabric topology is the tremendous scalability it
offers. As a company’s IT needs grow, the switched fabric can continue adding connections
as needed. Because each connection is a virtual point-to-point, this topology preserves high
bandwidth requirements throughout an enterprise. Switched fabric will also match speeds
between different speed connections, providing a high level of server versatility without
performance loss. Likewise, this type of connection allows the business to match cables
throughout a diverse network. The final advantage for a very large, very complex network
is the cost savings.
For smaller networks, however, the cost of switched fabric topologies can be a disad-
vantage. Latency becomes another problem when using switches. Though the ultimate
connections are clear and fast, there can often be some delay in making the connections
between nodes due to the complexity of the system. In addition, connections are non-
deterministic, which means that two successive connections between the same ports may
take an entirely different path each time with attendant differences in latency. The result is
that sequential commands can possibly arrive out-of-order at a peripheral. Some switches
offer an option to guarantee command ordering.
A switched fabric topology would generally be the best choice for very large enter-
prises that routinely handle high volumes of complex data between local and remote
servers. It provides remarkable scalability and f lexibility, while sustaining high bandwidth
requirements.
Loop TopologyLoop topology interconnects up to 126 servers, storage subsystems, or storage devices in a
ring. In an arbitrated loop, each server arbitrates for loop access and, once granted, has a
dedicated connection between sender and receiver. The available bandwidth of the loop
is shared among all servers. The primary reason to use a loop is cost. Because no switch
is required to interconnect multiple devices, the per-connection cost is significantly less
than with the switched fabric method. The peripherals in a loop need not worry about the
possible 16 million addresses in a switched environment. With only 126 possible addresses
to deal with, peripheral costs can also be reduced. Figure 3 shows the loop method.
Fibre Channel Technology 5
Figure 2.Switched Fabric
One variation on a straight loop configuration is the addition of hubs to the loop.
A hub is a hardware component with multiple ports that enables you to interconnect several
devices in the loop without breaking the loop. Using a hub will not increase the possible
number of servers or storage items in the ring, as the loop limit of 126 addresses remains
true for the hub (the hub is not normally an addressable unit). Though loop topology is
always in a ring shape, using hubs gives the ring more of a star-like shape, though it is still
a closed loop.
The reasons to use hubs in loop topology are that they allow non-disruptive expansion
to the network and allow removal-replacement of failed units while loop activity continues,
known as hot-plugging. Because hubs allow the mixing of copper and fiber cables, imple-
menting hubs in the loop allows greater distance between nodes so that the devices don’t
have to be as physically close to each other in the building. These less stringent distance
requirements and the simple configuration of hubs can help increase the f lexibility of loop
topology.
For the majority of small to medium networks, the advantages of a loop topology
outweigh any disadvantages. A loop is scalable from as few as 3 devices up to 126. Because
switches are not needed, this topology is quite cost-efficient for most enterprises. The loop
is best used to meet the scalability demands of disk storage configurations. It is in the most
demand today and will remain competitive in the future.
The disadvantages of a loop topology are most keenly felt by large networks. Loop is
rather inefficient in large configurations because only two nodes communicate at a time.
Also, large networks users can experience high latency because all the connections are
in a circle. Proposed enhancements to the Fibre Channel loop include a mode of operation
that eliminates arbitration and allows spatial reuse so that multiple “conversations” can
occur at the same time.
Another loop disadvantage is that all devices must operate at the same link speed.
Further, all devices in the loop must use the same cabling scheme, except when using
repeaters/translators or hubs. Basic physical loops have no bypass circuitry available during
power or cable interruptions, though using hubs in the loop will bypass this problem.
Finally, each additional node in the loop adds to latency.
Figure 3.Loop
6 White Paper
System Interface
FC-4
FC-3
FC-2
FC-1
FC-0
Upper Level Protocol
SCSI IPI-3 HIPPI Block MUX IP
Common Services
Framing Protocol/Flow Control
8B/10B Encode/Decode
266 Mbits/s 1062 Mbits/s531 Mbits/s
Despite these issues, loop connections are clearly the best choice for many network
cases. Loops are quite efficient when a network’s high bandwidth requirements are inter-
mittent. The topology is ideal for these situations. Likewise, when each node in the loop is
less than 100 meters apart, the loop topology is very effective.
How Is Fibre Channel Organized?
To allow each of the three unique topologies to work, Fibre Channel is organized into five
hierarchical functional layers and five service classes.
Functional LayersThe Fibre Channel structure is defined as a multi-layer stack of functional levels, not
unlike those used to represent network protocols. However, these layers do not map directly
to OSI layers. The five layers of the Fibre Channel standard define the physical media and
transmission rates, encoding scheme, framing protocol and f low control, common services,
and the upper-level application interfaces.
The five layers of Fibre Channel are shown in Figure 4. The user protocol transported
over Fibre Channel—such as SCSI or IP—is the upper-level protocol and is not touched by
the Fibre Channel layers.
The FC-0 layer defines the physical portions of the Fibre Channel, including the physi-
cal characteristics of the media, transmitters, receivers, and connectors that can be used; the
electrical and optical characteristics; the transmission rates; and other physical components
of the standard. This layer covers a wide range of performance and cost alternatives that
allow systems integrators to tailor installations to meet the specific customer needs.
The Fibre Channel FC-0 layer supports a variety of physical media to address different
cabling requirements. For short-distance connections, such as attaching nodes on the same
f loor of a building, coaxial and twisted pair cables are quite effective. To connect hubs or
switches between f loors or between buildings, optical fiber cables are the best choice.
Because copper and optical fiber can use the same protocols, switches can handle both
types of cable simultaneously.
The FC-1 layer defines the 8B/10B encoding/decoding scheme used to integrate the data
with the clock information required by serial transmission techniques. Fibre Channel uses
10 bits to represent each 8 bits of upper-level data. The 10-bit transmission characters are
converted back into 8-bit bytes at the receiver. Using 10 bits for each character provides
1,024 possible encoded values rather than the 256 that 8-bit characters would allow.
Figure 4.Fibre Channel’s Five Layers
Fibre Channel Technology 7
The FC-2 layer disassembles and reassembles data sequences. The layer defines some
primitive commands—ordered sets—to handle configuration management, error recovery,
and signaling between two ends of a link.
In managing a data transfer sequence, the FC-2 layer recognizes that all frames
belonging to a single transfer are uniquely identified by a sequential numbering system
from 0 through n. After a transfer, the receiver can then determine if all the frames were
received. If it recognizes that any are missing, it can determine which specific frame(s) was
not transmitted.
The FC-2 layer also manages the different classes of service, which are discussed in the
“Service Classes” section of this paper.
The FC-3 layer is intended to provide the common services required for features such
as striping to multiply bandwidth. The ANSI committee is still defining the functions in
this layer.
The FC-4 layer provides seamless integration of existing standards. This layer accom-
modates a number of other data communications protocols such as FDDI, HIPPI, SCSI-3,
and Internet Protocol (IP), as well as IBM Single Byte Command Code Set (SBCCS) of
their OEMI and EXCON Channels, Ethernet, and ATM.
As Fibre Channel is applied to other applications such as the crate-to-crate transfers
between memory buses like VME and Futurebus, many more application interfaces are
expected to be defined. In all, the Fibre Channel standard provides the capability to
support up to 255 types of protocols.
Service ClassesTo accommodate a wide range of communications needs, Fibre Channel defines five
different classes of service. Unlike FDDI and other ring-type local area networks, each
class can be used between as few as two points, and each class can be integrated within
the framework of other network protocols.
Class 1, a hard or circuit-switched connection, functions in much the same way as
today’s dedicated physical channels. When a host and device are linked using Class 1, that
path is not available to other hosts. When the time needed to make a connection is short
or data transmissions are long, Class 1 is an ideal link. Two interconnected clustered servers
could use a Class 1 hookup to ensure rapid and smooth communication.
Class 2 is a connectionless, frame-switched link that provides guaranteed delivery with
an acknowledgment of receipt. No delay is required to establish connections as in Class 1.
Also, no uncertainty exists as to whether or not delivery was achieved.
If delivery cannot be made because of congestion, a Busy frame is returned, and the
sender tries again. The sender knows to retransmit immediately without waiting for a long
time-out to expire.
As with traditional packet-switched systems, the path between the two devices is not
dedicated, allowing better use of the bandwidth of the link. Class 2 is ideal for data transfers
to and from a shared mass storage system physically located at some distance from several
individual workstations. With Class 2, however, there is a delay in the turnaround time
needed for the recipient to receive a packet and generate a response, and for the sender to
receive and understand the response on a packet-by-packet basis.
Class 3 is a connectionless service that allows data to be sent rapidly to multiple recipi-
ents, but without confirmation of receipt. This class is most practical when it takes a long
time to make a connection. By not providing confirmation, Class 3 service makes much
8 White Paper
better use of the available bandwidth by reducing overhead. However, if a single user’s
link is busy, the higher software layers must determine whether or not data has been lost.
If data is lost, then the software must retransmit the data. Class 3 service is standard for
loops because the configuration of the loop is known to all and a busy port should not
be encountered. The reduced overhead is also very attractive to peripheral designers for
whom cost is a primary concern.
Class 4 is a virtual connection service that establishes up to 256 virtual circuits to pro-
vide guaranteed fractional bandwidth service between communicating ports. The service
multiplexes frames at frame boundaries to or from one or more ports with acknowledg-
ment provided. This class is used by many video applications that need a specific bandwidth
to keep information f lowing to video screens.
Class 6 is a service that provides dedicated connections for reliable multicast. A Class
6 connection is requested by a port for one or more destination ports. Destination ports
return an acknowledgment to a multicast server, which returns an acknowledge to the con-
nection initiator, establishing a reliable multicast connection.
What Media Do I Use for PhysicalConnections?
What kinds of media are suitable for these different connections and what impact, if any,
will the media have on performance?
Fibre Channel is defined to enable systems integrators to select several speed and
distance combinations. Within each media class, the components can be selected to suit
the most important objective—economy or performance. At 100 MB/s, distance ranges
from 20 meters with electrical signals over television coaxial cables, to 10 kilometers with
long wave-length optical signals over single-mode cables.
Optical Fiber or Copper CablesOptical fiber cable offers several advantages over copper cable, though copper is still a
viable option for many connections. Optical fiber cable has less signal attenuation and
offers reliable connection over much greater distances. Copper cable is very cost efficient,
though it is susceptible to noise. Electric ground loops that interfere with signal reception
can be caused by several copper wire network links between computers. Because fiber optic
cable uses light as the medium for data signaling, it is immune to induced electrical signals
and impulses. Because glass is an electrical insulator, fiber optic links avoid network ground
loops.
Fibre Channel Technology 9
Attribute Copper Cable Optical Fiber Cable
Distance 2 to 20 meters 2 to 10,000 meters
Cable cost Lower at short distances Lower at long distances
Transceiver cost Lower than fiber Higher than copper
Installation cost Average Above average
Total cost Lower <30 meters Lower >30 meters
Electrical noise High Low
Volt immunity Low Non-existent
Ground loops High Non-existent
Single-Mode or Multimode Fiber Optic CableSingle-mode fiber optic cable is optical fiber with a small core diameter (9 µm) in which
only a single mode, the fundamental mode, is capable of propagation. (A mode is a discrete
optical wave.)
Multimode fiber optic cable is an optical wave guide whose core diameter is large
(50 or 62.5 µm) compared with the optical wavelength, and in which a large number of
modes are capable of propagation. Several hundred modes can propagate in multimode
fiber. The upper limit to the number of modes is determined by the core diameter. Single-
mode fiber cable is many times more expensive than multimode fiber cable. Therefore, it
is typically used only when distance requirements are extremely high.
Is Fibre Channel Being Used Successfully Today?
The Fibre Channel technology has already had a profound impact on many industries
and on many particular technology problems. Two such examples are the digital media
(video) industry and the network attached storage (NAS) technology.
Fibre Channel has successfully solved some daunting challenges in the area of video
editing. In today’s post-production studios, processing film and video images is typically
completed in multiple stages by a variety of people. Image processing equipment used in
one stage is generally separated from those at other stages to avoid confusion and rework.
In the past, users only had access to their local storage. When one group completed
all of its tasks on a video segment, that data had to be physically transported from one stage
of editing to the next. There was no way to share the editing stages between remote groups,
which kept each editor isolated and made data transportation cumbersome. The extremely
high bandwidth requirements of video editing made online data transfers highly risky or
even impossible.
The Fibre Channel standard of data transfer allows studios to break down the storage
and transfer barriers and allows multiple technicians to access a single, highly reliable disk
recorder. This storing, playing, editing, and caching of video becomes centralized, avoiding
the chance of error in the production process.
10 White Paper
Table 1.Attributes of Copperand Optical Fiber Cable
SANFC-AL(SCSI & IP)
Parallel SCSI
Storage Servers
LAN
10/100Base-T100VG
Switched EthernetToken Ring
FDDIATM
Gigabit Ethernet
Clients
Though this example is specific to one niche industry, the idea of many users accessing
a centralized data storage device is the very concept creating a boon in NAS technologies.
NAS capitalizes on the emerging trend in computing infrastructure to disaggregate comput-
er resources and functions. Whereas networks used to revolve around a single, “do-it-all”
server, modern networks have many specialized servers for different functions and processes
such as email, web connections, printing, and databases. But none of these is helping to
relieve the I/O bottleneck that the networks have had all along. Conversely, the I/O problem
is getting progressively worse as networks rely on faster servers, increased data volumes, and
now Internet growth. The next step in this evolution is to detach storage from the central-
ized servers and spin it off as a separate function like the others.
As storage begins to move into a separate network function, the I/O squeeze must
be addressed and solved for an NAS system to operate at peak efficiency. By using Fibre
Channel technology as the high-speed backbone, it becomes easy and extremely efficient
to implement both servers and storage in a storage area network (SAN) as a counterpart
and complement to the existing local area network. This SAN architecture allows storage
to grow incrementally in a building-block approach with independence and connectivity
to several servers. Expanding the Fibre Channel topology to add multiple loops or switched
fabric configurations results in almost limitless scalability.
By using a SAN, companies also relieve the host processor of its storage tasks, thereby
freeing up valuable processing cycles for application processing. SAN architecture doesn’t
share the server’s configuration constraints, such as the number of I/O channels supported,
the cost of expansion cabinets to handle additional slots, and a single host’s limit of storage
devices. A SAN lets a company scale its server and storage environments independent of one
another, while still protecting its complete IT investment.
Fibre Channel is the enabling technology that successfully creates a networked storage
device that is available to all users. The high-speed data transfers and high bandwidth capa-
bilities make it more feasible than ever to store and access massive volumes of data. Users
can have 24-hour data availability even if they are remote. NAS is a Fibre Channel success
story that is still growing each day.
Fibre Channel Technology 11
Figure 5.The SAN Solution
What is the Latest Word on Fibre Channel?
Today’s customers are interested in finding and returning their data instantly, and that
data is larger than ever before because of the tremendous growth in video, multimedia,
and other large files.
Fibre Channel technology has brought to the market the high-performance, low-
cost communications pathway required by the intensive applications needed in today’s
networks. Fibre Channel also solves the distance, connectivity, and performance problems
seen in current systems using parallel SCSI.
The ability of Fibre Channel to support storage interconnect using SCSI commands,
data, and status—as well as transport network traffic—is providing entry to the mass
market. The resulting large sales volumes promise to drastically reduce costs, making it
possible that Fibre Channel will be the preferred interface for system interconnection. Fibre
Channel enables computing systems to transfer data at a rate of over 1 gigabaud per second,
or 10 to 250 times faster than today’s network protocols. Fibre Channel is positioned to be
the system interconnection means of the future.
12 White Paper
Appendix A: Fibre Channel Quick Reference Card
Project DocumentAcronym Title Number Number
FC-PH Fibre Channel Physical Interface 755-M X3.230-1994X3.230/AM1:1996
FC-PH-2 2nd-Generation Physical Interface 901-D X3.297-199x
FC-PH-3 3rd-Generation Physical Interface 1119-D
FC-AL Arbitrated Loop 960-D X3.272-1996
FC-AL-2 2nd-Generation Arbitrated Loop 1133-D
FC-FG Generic Fabric Requirements 958-D X3.289-199x
FC-SW Switch Fabric 959-D
FC-GS Generic Services 1050-D X3.288-199x
FC-GS-2 2nd-Generation Generic Services 1134-D
FC-AE Avionics Environment 2009-D
FC-AV Audio-Visual Transport Proposed
FC-BB Backbone Proposed
FC-FP Mapping of HIPPI-FP 954-M X3.254-1994
FC-LE Link Encapsulation 955-D X3.287-199x
FC-SB Mapping of Single-Byte 957-M X3.271-1996Command Code Set
SCSI-FCP SCSI-3 FC Protocol 993-R X3.269-1996
SCSI-GPP Generic Packetized Protocol 991-DT X3/TR-16-1995
FC-I3 Revision to IPI-3 Disk standard 496-R X3.291-199x
FC-I3 Revision to IPI-3 Tape standard 505-R X3.290-199x
HIPPI-FC Fibre Channel Frames on HIPPI 979-D X3.283-1996
FC-FLA Fabric Loop Attachment 1235-DT
FC-PLDA Private Loop Direct Attach 1162-DT
FC-SL Slotted Loop 1232-D
10 BIT 10-Bit Interface Technical Report 1118-DT X3/TR-18-199x
FC-CU Cu Interface Implementation Guide 1135-D
Jitter Methodology of Jitter Specification 1230-DT
DocumentOriginator Title Rev. Number
FCSI Common FC-PH Feature Sets 3.1 FCSI-101
FCSI SCSI Profile 2.2 FCSI-201
FCSI IP Profile 2.1 FCSI-202
FCSI Generic Link Module 1.0 FCSI-301
FCA N_Port to F_Port Interoperability 1.0
FCA IP Profile In development
FCA Hub Profile In development
Fibre Channel Technology 13
Table A1.Fibre Channel andRelated Standards and Technical Reports
Table A2.FCA and FCSI Profilesand Related Documents
31 Byte 0 24 23 Byte 1 16 15 Byte 2 8 7 Byte 3 0
Start of Frame (SOF)
R_CTL Destination Identifier (D_ID)
CS_CTL Source Identifier (S_ID)
TYPE Frame Control (F_CTL)
SEQ_ID DF_CTL Sequence Count (SEQ_CNT)
Originator Exchange Identifier (OX_ID) Responder Exchange
Identifier (RX_ID)
Parameter
Reserved for Extended Frame Header (16 Bytes)
Reclaimed from Expiration Security Header Reserved (16 Bytes)
D_NAA Network Destination Address (high order bits)
Network Destination Address (low order bits)
S_NAA Network Source Address (high order bits)
Network Source Address (low order bits)
Validity Originator Process Associator (most significant 3 Bytes)
Originator Process Associator (least significant 4 Bytes)
Reserved Responder Process Associator (most significant 3 Bytes)
Responder Process Associator (least significant 4 Bytes)
Originator Operation Associator (most significant word)
Originator Operation Associator (least significant word)
Responder Operation Associator (most significant word)
Responder Operation Associator (least significant word)
Device Header of 16, 32, or 64 Bytes(For use of protocol mappings)
Payload Data of0 to (2112 minus length of optional headers) Bytes
CRC
End of Frame (EOF)
Bit Function Bit Function Bit Function
Extended Frame Hdr. Network Header Device Header
23 0 Absent 21 0 Absent 00 None17
1 Present (Reserved) 1 Present 01 16 Byte
Reclaimed Association Header to 10 32 Byte
22 Expiration/Security 20 0 Absent 16 11 64 Byte
Header 1 Present
14 White Paper
Table A3.FC-PH Frame
Table A4.DF_CTL Bitsfor Data Frames
Field Non-Basic Basichex Data Frame Link Service Link Service Link Control
R_CTLbits 31-28 0 FC-4 Device 2 Extended 8 C
(# in filed ➡) 4 Video 3 FC-4(Routing)
0 Uncategorized NOP ACK_1
1 Solicited Data Reserved ABTS ACK_NACK_0
2 Unsolicited Control Request RMC P_RJT
R_CTL 3 Solicited Control Reply Reserved F_RJT
bits 4 Unsolicited Data BA_ACC P_BSY
27-24 5 Data Descriptor BA_RJT F_BSY(Data)
Info 6 Unsolicited Reserved F_BSYCategory Command (Link Control)
7 Command Status Reserved LCR
8 NTY
9 Unspecified END
Others Reserved
Type FC-4 Device 1z F_BSY(# in filed ➡) see next page 1 Extended 0 3z P_BSY
Video reserved 3 FC-4 (z-see note)others reserved
0 Abort, discard multiple Seq.
F_CTL 1 Abort, discard single Seq. Not See next
bits 5-4 2 Process policy, with ∞ buffers Meaningful page
3 Discard multiple Seq with retransmit
DF_CTL All valid None Not Meaningful
ACK, 0, 1, NParameter Relative Offset Not Meaningful BSY, RJT
(Optional) reason & action
Payload 0-2112 Bytes 0-12 Bytes 0 Byte
Notes: z in Type field is the information category of frame being busied
Bit Function Bit Function Bit Function
Simplex Camp-ON
31 0 Duplex 29 0 Not Requested31
1 Simplex 1 Requested Virtual Circuit
Stacked Connect Buffered Class 1 to Identifier
30 0 Not Requested 28 0 Not Requested 24
1 Requested 1 Requested
Fibre Channel Technology 15
Table A5.Header Fieldsby Frame Group
Table A6.CS_CTL Bitsfor All Frames
31 Byte 0 24 23 Byte 1 16 15 Byte 2 8 7 Byte 3 0
Action code Reason Code Reserved VendorP_BSY and F_BSY P_BSY Unique01 Seq terminated 01 Physical busy02 Seq active 03 Resource busyOthers reserved FF Vendor Unique
F_BSY01 Fabric busy03 Nx_Port busyOthers reserved
P_RJT and F_RJT P_RJT and F_RJT01 Retry see next page02 Non-retryableOthers reserved
hex FC-4 hex FC-4 hex FC-4
04 LLC/SNAP (in order) 16 CP IPI-3 Slave 24 SNMP
05 LLC/SNAP 17 CP IPI-3 Peer 40 HIIPI-FP
08 SCSI-FCP 19 SBCCS-Chan 34 to MessageWay
09 SCSI-GPP 1A SBCCS-CU 37
11 IPI-3 Master 20 FC Services 5D Fabric Controller
12 IPI-3 Slave 21 FC-FG E0
13 IPI-3 Peer 22 FC-SW to Vendor Unique
15 CP IPI-3 Master 23 FC-AL FF
Table A7.Parameter Field for BSY and RJT
Table A8.“Device” Type FieldValues of FC-4 Frames
16 White Paper
Bit Function Bit Function Bit Function
Exchange Context X_ID Reassigned23 0 Originator 15 0 Retained Continue Seq.
1 Responder 1 Reassigned 7 00 No information
Seq. Context Invalidate X_ID to 01 Immediate22 0 Initiator 14 0 Retained 6 10 Soon
1 Recipient 1 invalidated 11 Delayed
First Sequence ACK form Abort Sequence21 0 Last or middle 13 00 No assistance 5 ACK frame
1 First sequence to 01 ACK_1 req. to 00 Continue Seq.
Last Sequence 12 10 ACK_N req. 4 01 Abort sequence20 0 First or middle 11 ACK_0 req. 10 Stop sequence
1 Last sequence 11 retransmit seqData frame (prev pg)
End Sequence Data Compression Relative Offset19 0 Cont. Sequence 11 0 Uncompressed 3 0 Not meaningful
1 Last Data Frame 1 Compressed 1 Present
End Connection Data Encryption Reserved18 0 Retain 10 0 Unencrypted 2 Exchange
1 Disconnect/Deactivate 1 Encrypted reassembly
Reclaimed Retransmit Seq. Fill Data Bytes17 Chained sequence 9 0 Original 1 00 None
1 Retransmitted to 01 One Byte
Seq. Initiative Unidirect/Remove 0 10 Two Byte16 0 Hold 8 0 Bidirect/Deactivate 11 Three Byte
1 Transfer 1 Unidirect/Remove
Code Function By Code Function By Code Function By
01 Invalid D_ID F,P 0D Invalid SEQ_ID P 19 Expiration/ PSecurity error
02 Invalid S_ID F,P 0E Invalid DF_CTL P 1A Fabric path Fnot available
03 Nx_Port F 0F Invalid P 1B Invalid VC_ID Fnot avail (temp) SEQ_CNT
04 Nx_Port F 10 Invalid P 1C Invalid F,Pnot avail (perm) Parameter field CS_CTL field
05 Class not F,P 11 Exchange error P 1D Unable to F,Psupported establish VC
06 Delimiter usage F,P 12 Protocol error P 1E Dedicated F,Perror Simplex error
07 Type not F,P 13 Incorrect length F,P 1F Invalid class F,Psupported of service
08 Invalid Link P 14 Unexpected P 20 Pre-emption FControl ACKP rejected
09 Invalid R_CTL P 15 Class invalid F 21 Pre-emption Ffield at FFFFFE not enabled
0A Invalid F_CTL P 16 Log-in required F,P 22 Multicast Ffield error
0B Invalid OX_ID P 17 Excessive Seq. P 23 Multicast Fattempted error term
0C Invalid RX_ID P 18 Establish P FF Vendor PExchange error unique error
Table A9.F_CTL Bits for AllFrames (Word 2)
Table A10.Reject (P_RJT, F_RJT) Reason Codes
Fibre Channel Technology 17
Optical Performance Dist.Transmitter Media MBaud Mbytes/s km
Longwave LED 62.5µm 132.8125 12.5 1.5Multimode Fiber
Longwave LASER Single-mode Fiber 2 or 10
Longwave LED 62.5µm 265.625 25 1.5Multimode Fiber
Longwave LED 50µm 1.5Multimode Fiber
Shortwave LASER 50µm 2(with OFC) Multimode Fiber
Shortwave LASER 50µm 1(with OFC) Multimode Fiber 531.25 50
Longwave LASER Single-mode Fiber 10
Shortwave LASER 50µm 0.5(with OFC) Multimode Fiber
Shortwave LASER 50µm(w/o OFC) Multimode Fiber 1062.5 100 0.5
Longwave LASER Single-mode Fiber 2 or 10
Shortwave LASER 50µm 0.3(w/o OFC) Multimode Fiber 2125 200
Longwave LASER Single-mode Fiber 2
Shortwave LASER 50µm 0.175(w/o OFC) Multimode Fiber 4250 400
Longwave LASER Single-mode Fiber 2
Table A11.Optical Feedsand Speeds
18 White Paper
Electrical Performance Dist.Transmitter Media MBaud Mbytes/s km
Long Video Coax (RG-6) 67
ECL Video Coax (RG-59) 56
Intra-cabinet Mini Coax (RG-179) 16
Shielded twisted pair (STP-1) 32
Twinax 37
Long Video Coax (RG-6) 132.8125 12.5 100
ECL Video Coax (RG-59) 100
Inter-cabinet Mini Coax (RG-179) 42
Shielded twisted pair (STP-1) 80
Twinax 93
Long Video Coax (RG-6) 47
ECL Video Coax (RG-59) 40
Intra-cabinet Mini Coax (RG-179) 11
Shielded twisted pair (STP-1) 22
Twinax 26
Long Video Coax (RG-6) 265.625 25 100
ECL Video Coax (RG-59) 100
Inter-cabinet Mini Coax (RG-179) 28
Shielded twisted pair (STP-1) 57
Twinax 66
Long Video Coax (RG-6) 33
ECL Video Coax (RG-59) 28
Intra-cabinet Mini Coax (RG-179) 7.6
Shielded twisted pair (STP-1) 16
Twinax 18
Long Video Coax (RG-6) 531.25 50 84
ECL Video Coax (RG-59) 71
Inter-cabinet Mini Coax (RG-179) 19
Shielded twisted pair (STP-1) 40
Twinax 46
Long Video Coax (RG-6) 24
ECL Video Coax (RG-59) 20
Intra-cabinet Mini Coax (RG-179) 5.6
Shielded twisted pair (STP-1) 11
Twinax 13
Long Video Coax (RG-6) 1062.5 100 59
ECL Video Coax (RG-59) 50
Inter-cabinet Mini Coax (RG-179) 14
Shielded twisted pair (STP-1) 28
Twinax 33
Table A12.Electrical Feedsand Speeds
Fibre Channel Technology 19
Characteristic Class 1 Class 2 Class 3 Class 4 Class 6
Dedicated Multiplexed Fractional UnidirectionalFunction Connection (Frame Datagram Guaranteed Dedicated
Switched) Bandwidth Connection
Fabric discards No No Yes No Noframes?
Communications One to One Many to Many One to Manytype btw. ports
Initial round Yes No No No Yestrip delay
Hunt group support Optional Optional Optional Optional Optional
Multicast support No No Optional No Yes
Broadcast support No No Optional No No
Stacked ConnectRequest
Dedicated Simplex
Camp-On Optional Not applicable Optional
Buffered Class 1
Unidirectional YesConnection
End-to-End Yes Yes No Yes Yesflow control
Buffer-to-Buffer Yes Yes Yesflow control
Yes, only for Yes, only forFabric can reject Only undeliverable invalid Only
frames? Connection frame activation Connectionrequest No, frame is frame request
Fabric busies frame Yes, for long- discarded frameframes? term resource No
contention
Delivery Order Yes Depends on Fabric, yes Yes Yesguaranteed? for most fabrics
Nx_Port supports Optional Optional Optional Optional, No OptionalClass of Service? for NL_Port
Fabric supports Optional Optional Optional Optional OptionalClass of Service
Topics to be included in future enhancements:
• 8 and 16 GBaud physical variants
• Asynchronous and Isochronous services
• Striping
• Mapping for many more existing and new protocols
Table A13.ClassCharacteristics
20 White Paper
Address Function Address Function
FFFFFF Broadcast Alias_ID FFFFF9 Quality of Service Facilitator
FFFFFE Fabric F_Port FFFFF8 Alias Server
FFFFFD Fabric Controller FFFFF7 Security Key Distribution Server
FFFFFC Directory Server FFFFF6 Clock Synchronization Server
FFFFFB Timer Server FFFFF5 Class 6 Multicast Server
FFFFFA Management Server 000000 Unidentified
Unisys is a registered trademark of Unisys Corporation. Intel is a registeredtrademark of Intel Corporation. Compaq is a registered trademark of CompaqComputer Corporation. Hewlett Packard is a registered trademark of Hewlett-Packard Company. Tektronix is a registered trademark of Tektronix, Inc. Sun is aregistered trademark of Sun Microsystems, Inc. Intel is a registered trademark ofIntel Corporation. Symbios Logic is a trademark of Symbios Logic, Inc. NEC is aregistered trademark of NEC Electronics, Inc. Digital is a registered trademark ofDigital Equipment Corporation. Microsoft and Windows NT are registered trade-marks of Microsoft Corporation. Other brand names, product names and trade-marks are acknowledged to be the property of their respective owners.
© 1997 Unisys Corporation.
Fibre Channel Technology 21
Table A14.Well-Known Address Identifiers
@@@@@@@@e?@@@@@@@@e?@@h?@@h?@@h?@@h?@@h?@@h?
@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e
@@@@@@@@@@@@@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
?@@?@@?@@?@@?@@?@@
?@@@@@@@@?@@@@@@@@
?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@e?@@@@@@@@?e@@@@@@@@
@@g@@g@@g@@g@@g@@g@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@
WHEN INFORMATION IS EVERYTHING
Unisys is one of a select group of companies with the broad portfolio of ser-vices, technologies and third-party alliances needed to deliver the benefits ofinformation management—helping clients use their information assets toenhance their competitiveness and responsiveness to customers.
Our expertise in information management is founded on the strengths of ourthree global businesses: consulting, outsourcing, solutions and systems inte-gration; industry-leading technologies; and comprehensive services and prod-ucts supporting distributed computing environments.
Printed in U S America 9/97