Next Generation SONET/SDH

Chapter 3.6 - Fiber Channel


A channel is a well-defined, direct, and structured mechanism to transport information
data between a source and a destination, and in some cases, to several destinations.
Most of the decision making takes place before and during setting up a channel.
When a channel is set up, very few decisions are required and made. In the data
space, a channel is set up to connect peripheral devices from a workstation to a peripheral
device such as a printer.

Channels are specified by common channel protocols and for a limited distance
between source and destination, typically a few feet or meters. Although short distances
may be adequate for small systems, they are not for large systems connected
with large-capacity peripherals at long distances. This is the application space of the
Fiber Channel (FC) protocol.

The Fiber Channel was first developed for high-performance devices communicating
with processors and for intercommunication between many processors. The
FC physical interface defined a system that consists of a set of a general-purpose
processors, servers, or subsystems. A node may contain one or more node_ports. A
node_port (or N_Port) is the function that connects the node with another node or to
the switching fabric. As such, FC was specified to support several transmission media:
coaxial, shielded twisted copper pair with a DB-9 connector, multimode fiber
(62.5 and 50 μm core), and single-mode fiber (9 μm core). LED or lasers may be
used at wavelengths 780 and 1300 nm. The maximum distance is 500 m to 10 km
for optical transmission and 10–100 m for electrical.

It is also specified to support several bit rates. The most common rate is at 1,063
Mbaud, also termed “full speed.” This rate is derived from an initial 100 Mbytes/s
(10X) and adding 63 Mbit/s overhead: 100 × 10 + 63 Mbit/s = 1,063 Mbit/s. In addition,
there are multiple rates (2× at 2,126 Mbit/s, and 4× at 4,252 Mbit/s) and also
subrates (1/2×, 1/4×, and 1/8×, at 50, 25, and 12.25 Mbytes/s, respectively).

Unlike the open-system interconnect (OSI) that defines seven layers, the Fiber
Channel Protocol defines five. These five layers are named FC-0 to FC-4. The first
three (F0 to F2) are the responsibility of the N_Port.

  • FC-0 is equivalent to the link layer. It specifies the physical interface—media,
    receiver, transmitter, and signaling—the medium, and data rate. For example,
    at 100 Mbytes/s, the I/O port may support 1300 nm laser over SMF, 780 nm
    laser over 50 μm MMF, or electrical over coaxial. In contrast, at 12.5
    Mbytes/s it may support 1300 nm LED over 62.5 MMF, electrical over coaxial,
    or shielded twisted pair.
  • FC-1 is the transmission protocol that specifies the link maintenance aspects
    and it encodes 8 bit EBCDIC binary codes to unique 10 bit codes (8B/10B encoding/
    decoding; see above).
  • FC-2 is the signaling protocol that specifies the frame format, segmentation
    and reassembly, flow control, classes of services, exchange and sequence
    management, topologies, and login/logout procedures.
  • FC-3 specifies the common services for multiple ports in one mode; it is reserved
    for future functions.
  • FC-4 specifies the upper layer protocol (ULP), and it is responsible of mapping
    a variety of well-known higher layer protocols, such as:
    —Internet protocol (IP)
    —Asynchronous transfer mode (ATM) using adaptation layer 5 (AAL-5)
    —Small computer system interface (SCSI)
    —High-performance parallel interface (HIPPI)
    —Intelligent peripherals interface-3 (IPI-3)
    —Single byte command code sets (SBCCS)
    —Fiber connectivity (FICON)
    —Enterprise system connection (ESCON)
    —Video/audio multimedia and networks

The Fiber Channel frame is preceded and concluded by idle frames to provide margin
between frames. The structure of the FC frame is illustrated in Figure 3.10.

FC defines a port that consists of a transmitter with outbound traffic and a receiver
with inbound traffic (this definition is applicable to most communications
ports). However, with this in mind, FC is suited to two key topologies: the point-to-point
and the arbitrated loop (Figure 3.11).

Figure 3.10 FC protocol frame structure. The numbers in parenthesis are in octets; however, after 8B/10B encoding, they become 10-bit words.

The point-to-point topology requires a two-fiber link, one for transmit and the
other for the receive direction. This topology requires only a simple link initialization
before communication begins.

The arbitrated loop can connect up to 127 ports in a single network. The media is
shared among many devices, thus limiting each device’s access to the network. In
this topology, devices must support the Arbitrated Loop Initialization protocol before
communication begins.

In addition, FC supports a fabric switch topology and a fabric switch with arbitrated
loops topology (Figure 3.12), which can potentially interconnect 224 devices
and allow for simultaneous communication.* When FC ports log-in the fabric, the

*The fabric switch is nonblocking. The I/O ports of the FC fabric are known as F_Ports.

Figure 3.11 FC supports two distinct topologies: a point-to-point topology (A) and an arbitrated loop topology (B).

Figure 3.12 An FC fabric switch also supports the two topologies. (A) The fabric switch with point-to-point connectivity. (B) The fabric switch with arbitrated loop topology capable to interconnect 224 devices and allow for simultaneous communication.

fabric assigns native address identifiers to them. Thus, the fabric may be a broadcast
server, a directory server, a multicast server, an alias server, and so on.

To avoid congestion, FC uses a credit-based flow control scheme. Each transmitting
node is given a number of credit tokens according to buffer size. When transmission
of frames begins, these credits are decremented. Frame transmission is
halted if there are no credits left. Upon receiving acknowledgment, the credits are
replenished and transmission may start again. Thus, this predictable credit schemes
guarantees that no frames are lost due to buffer overflow.

During the loop initialization process (LIP), each port is dynamically assigned
an 8-bit Arbitrated Loop Physical Address (AL_PA). Although AL_PA is 8 bits, up
to 127 addresses are assigned.

  • The initialization process begins with transmission of a LIP Primitive
    Sequence upon power-on of a port or when a loop failure is detected. As the
    port starts transmitting a LIP, it triggers other ports on the loop to transmit
    a LIP.
  • During this phase, a master port is selected, either the one with the lowest numerical
    port name, or one that has a fabric switch.
  • The next step is to allow a port to select an AL_PA. The loop master transmits
    a 127-bit bitmap of a frame around the loop and each port seizes a bit. There
    are four priorities assigned, the highest, known as loop initialization fabric assigned
    (LIFA) goes to the fabric switch. The other three are loop initialization
    previously assigned (LIPA), loop initialization hard assigned (LIHA), and
    loop initialization soft assigned (LISA). Thus, by the time the bitmap frame
    has come back to the master, all ports on the loop have been assigned an
  • Finally, the master transmits a primitive signal to indicate that the initialization
    process is complete.



Already an Engineering360 user? Log in.

This is embarrasing...

An error occurred while processing the form. Please try again in a few minutes.

Customize Your Engineering360 Experience

Category: Fiber Optic Couplers
Privacy Policy

This is embarrasing...

An error occurred while processing the form. Please try again in a few minutes.