This is a United States Navy Computer. See
Welcome - FDDI Frequently Asked Questions (FAQ)
This FDDI page is currently under construction and you can
expect changes frequently.
Revision 4 January 4, 1996
(Bob Grow, Scott Hiles)
Revision 3 December 14, 1995
(Bob Grow, Scott Hiles)
Revision 2 September 7, 1994
Revision 1 Unknown
Your input on
changes/corrections is welcome. Please include an explanation of what is
correct if you find something incorrect. Comments and corrections to:
These answers are courtesy of (last known addresses):
and others, who's identity has been lost
in the multiple ownership of the FAQ.
It is a 100 Mbps Local Area Network, defined by ANSI and OSI standards.
It was originally designed to operate over fiber optic cabling, but now
also includes standard copper media for interconnection. FDDI uses a
token ring media access control protocol.
Most of the standardization of FDDI was done in Accredited Standards
Committee X3T9.5. In 1995, X3T9.5 became X3T12. FDDI standards are
approved by both ANSI (American National Standards Institute) and ISO
(International Standards Organization). ISO approval usually occurs
after ANSI approval.
A token is a special three octet FDDI frame.
The station waits until a token comes by, grabs the token,
transmits one or more frames and release the token. The amount
of frames that can be transmitted is determined by timers in
the MAC protocol chips.
FDDI uses a timed token protocol, while 802.5 uses a
priority/reservation token access method. Therefore, there are
some differences in frame formats, and significant differences in
how a station's traffic is handled (queueing, etc.) Management of
the rings is also very different.
When a DAS is connected to two concentrator ports, it is called
dual-homing. One port is the active link, where data is
transmitted and the other port is a hot standby. The hot standby
link is constantly tested and will kick in if the active link
fails or is disconnected. The B-port in a DAS is usually the
active port and the A-port the hot-standby.
DAS (Dual Attachment Station) is a station with two peer ports (A-Port
and B-Port). The A-port connects to the B-Port of another DAS,
and the B-port is going to connect to the A-Port the yet another
When a link in the dual-ring is broken or not connected, the two
adjacent ports connecting to the broken link will be removed from
the ring and the both stations enter the wrap state. This joins
the two counter-rotating rings into one ring.
Usually you will need a concentrator port (M-Port) to connect
each SAS. A DAS can be connected in the dual rings or to
concentrator port(s). FDDI allows two S ports to be connected
(a two station ring), or an S port to be connected to an A or
B port of a DAS causing a wrapped dual ring. If more than one DAS
is used, ring redundancy is lost. (Not all equipment vendors
allow S to A, B, or S connections without special configuration.)
Advantages: Fault tolerance. When a link breaks, the ring
can be segmented. A concentrator can just bypass the problem
port and avoid most segmentations. It also gives you better
physical planning. Usually people prefer tree physical
topology. Generally star configuration of a concentrator system
is easier to troubleshoot. Stations can be powered off without
serious ring effects.
Disadvantages: A concentrator represents a single point of failure.
A concentrator configuration may also be more costly.
A bypass relay is a device that is used to skip a station
on the dual ring if it is turned off, without causing the ring to
wrap. One problem with them is that they attenuate the light in
the fiber, so you can't have too many of them. (The maximum
number in a row depends on the bypass loss, and how the cable
plant is constructed. When the bypass joins two fiber links,
the number of connectors between the optical transmitter and receiver
will usually increase.)
All the FDDI optical PMDs use the same wavelength (1300nm), so
they can be connected together. For example, PMD can be connected
to LCF-PMD as long as you stick to LCF-PMD configuration rules.
If you don't understand this optical stuff, don't attempt to
mix SMF-PMD with PMD or LCF-PMD devices without advice.
Too much optical power can permanently damage someone's eyes.
This is not generally a problem with PMD and LCF PMD but can be
with SMF-PMD especially Category II SMF-PMD. Inspecting the end
of any fiber (e.g., with magnification) without knowing what is at
the other end is not a smart thing to do.
Any problems that arise are generally with the transport
protocol. Frame fragmentation is standardized for TCP/IP. It
should also be noted that frame fragmentation will not work for
DECNET, IPX, LAT, Appletalk, NETBEUI etc. IP is the only protocol
that has a standard method of fragmenting. Other protocols
destined for Ethernet LANs must stay below the 1500 MTU.
FDDI frames have a max size of 4500 bytes and Ethernet only 1500
bytes. Therefore your bridge or router needs to be smart enough
to fragment the packets (e.g., into smaller IP fragments). Or you
need to reduce your frame size to 1500 bytes (of data).
(8 or more Idle symbol pairs at initial transmission)
Starting Delimiter (JK)
(J followed by K control symbol)
Frame Control (nn)
(Tell you if it is a token, MAC frame, LLC frame,
SMT frame, frame priority, sync or async)
Destination Address (nn)
(6 bytes of MAC Address in MSb first format)
Source Address (nn)
(6 bytes of MAC Address of this station)
Information field (nn)
(Variable Length. Usually starts with LLC header,
then SNAP field, then the payload e.g., IP packet)
Ending Delimiter (T)
(one T control symbol)
Frame Status (EAC)
(Three symbols of status: Error, Address_match,
and Copied. Each symbol is either SET or RESET.
e.g., If EAC == RSS, then the frame has no error,
some station on the ring matched the DA, and some
station on the ring copied the frame into its buffer.
Depends. You can get aggregate usage greater than 95 Mbps in
most installations. As with any LAN, high utilization corresponds
to longer queuing times. Many systems run fine at 75 Mbps. The
maximum for a station is implementation dependent. Some can't do
more than 20 Mbps, while others can sustain more than 90 Mbps.
After the buffer fills Frames start dropping. This is not a
problem unique to FDDI however. Consider Ethernet to T1, or
multiple Ethernets to a single Ethernet, or a lightly loaded
Ethernet to a heavily loaded Ethernet.
FDDI protocol analyzers of varying capability are available from
multiple manufacturers. Many FDDI equipment vendors also have management
software that implements ring mapping, inspection of other stations
operational parameters, general ring state monitoring, etc.
An optical time domain reflectometer (OTDR) and optical power meters
are used for testing optical fiber. There are also FDDI link testers
that measure power and low level FDDI protocol response.
Beacon is a special frame that FDDI MAC sends when something is
wrong. When Beaconing persists, SMT will kick in to detect and
try to solve the problem. A few FDDI implementations will beacon
on initial entry into the ring, but this is a short term
Just like any networking products, Ethernet, Token, FDDI, ATM,
there is a possibility that one vendor does not work with another.
But most of the equipment shipping today is tested for
interoperability. There are test labs like UNH and ANTC. Ask the
vendor what type of testing they did.
MaxTime = ~(#of stations * T_neg)
(T_neg is the negotiated target token rotation time)
Even at maximum load, this is unlikely. To get this worst case the
ring basically must go from zero offered load to maximum offered load
(all stations enqueueing ~T-neg of frames) virtually instantenously.
If this doesn't happen, the ~T_neg of available bandwidth is split
between multiple stations with each fraction of the bandwidth being
passed along the ring with each token rotation. This results in a given
station having multiple transmit opportunities within MaxTime.
To reduce MaxTime change the T_req of a station to some lower value
(e.g., 8 msec). This is done through the MIB parameter fddiPATHMaxT-Req.
Basically anything will be at least a bit faster. From NFS to
images transmission. Even if a single station cannot take advantage
of the 100 Mbps, the aggregate bandwidth will help a lot if
your Ethernet is saturated. However, note that though FDDI has higher
bandwidth than Ethernet, the signals travel at the same speed.
The propagation of a signal on the transmission line is the same for
Ethernet, token ring, and FDDI.
Depends. Let's do SAS first, it is easier. If a SAS is connected to
a concentrator, then the concentrator will bypass the SAS connection
using an internal data path. If a DAS is connected to a concentrator,
then the concentrator will also bypass the DAS. If a DAS is connected
to the trunk rings without using an optical bypass switch, then the
trunk ring will wrap. If multiple stations power off on the trunk
rings, then the ring will be segmented. Now if a DAS is using an
optical bypass switch, the switch will kick in and prevent the ring
If either conductor (fiber or wire) in a link is broken or disconnected
or loses a transmitter or receiver the link is removed from the ring.
SAS connecting to concentrator:
Same as above.
DAS dual-homed to concentrator(s):
If A-port link breaks, usually no effect since A port is generally
the backup port. (And SMT will NOT send out alert msg.)
If B-port fiber breaks, A-port will be joined into the ring after a
DAS on trunk rings, with no bypass:
If one link breaks, then the ring will wrap.
If both links break, ring will wrap, station won't be able to
DAS on trunk rings, with bypass:
When a bypass is switched it breaks its connections before making the
new ones. This causes a temporary wrap when either inserting or
removing a station from the ring. The A and B ports of the inserting
station will usually be added serially, though it may appear to be
If any one of the four link entering the bypass breaks, the ring will
wrap, living the bypass station in the ring.
If both links between the bypass and its neighboring stations break,
or both links between the bypass and its station break, the ring will
wrap, and the station connected to the bypass won't be able to
Connect concentrators and other equipment which is powered on all the
time (e.g., bridges, routers, ring monitors, etc.) into the trunk ring.
In a large enterprise, workgroup concentrators and users stations are
then connected to the backbone concentrators. Some connect bridges and
routers and critical servers to backbone concentrators using dual-
Graceful Insertion is a method to insert a station (or a tree)
into a ring minimizing disruption. Graceful Insertion is not
standard, and therefore different by vendor. Some implementations are
both "frame friendly" (don't corrupt frames) and "token friendly" (don't
destroy the token). The theory goes that Graceful Insertion can
minimize ring non_op and lost frames, therefore saving transmission
timeout in upper layer protocols (e.g., TCP).
The following is the counter argument: Graceful Insertion can use more
ring bandwidth (holding the token) than is consumed by a ring recovery.
And upper layer protocols are designed to perform frame recovery and
retransmission anyway. Also, no vendor can guarantee 100% Graceful
Station Management (SMT). It is part of the ANSI FDDI Standards
that provides link-level management for FDDI. SMT is a low-level
protocol that addresses the management of FDDI functions provided
by the MAC, PHY, and PMD. It performs functions like ring recovery,
frame level management, link control, etc. Every stations on
FDDI needs to have SMT.
A lot of FDDI equipment was shipped before the SMT standard was
completed. Most of this equipment was shipped with SMT software based
on SMT Revision 6.2, and earlier shipments were even made with SMT 5.1.
SMT 7.3 was the final working document in the standards committee. It
is functionally identical to SMT 7.2 and the approved SMT standard.
All of these versions of SMT work together on a ring, but they will look
different to a management station. (The SMT 6.2 and SMT 7.2 MIBs are
very different, as are the frame protocols related to the MIB.)
A port is the connector and supporting logic for one end of an FDDI
link. Each port has a transmitter and a receiver. Ports are given
names descriptive of their position in FDDI topologies. SMT defines
four types of ports (A, B, M, S). A dual-attachment station has two
ports, one A-port and one B-port. A single attach station has only
one port (S-port). A concentrator will have many M-port for connecting
to other stations' A, B or S-ports.
When connecting DASs, one should connect the A-port of one
station to the B-port of another. S-port on the SAS is to
connect to the M-port on a concentrator. A and B-port on
DASs can also connect to the M-port of a concentrator. M-ports
of concentrators will not connect to each other.
In more detail, SMT suggests the following rules:
V indicates a valid connection
I indicates an illegal connection
U indicates an undesirable connection with notification to SMT required
W indicates, if Active, prevent THRU in CFM and Port B takes precedence