FDDI

Home - Network - Lec1&2 - Lec3  - Lec3b - Lec4 - Lec5 &6 - Lec7  - Lec8 - Lec9  - Lec11 - Lec12


One of the older, and more robust, of the LAN technologies is the Fiber Distributed Data Interface (FDDI). FDDI was standardized via the ANSI X3T9.5 specification during the mid 1980s. At that time, high performance UNIX workstations were beginning to appear. These workstations needed a higher performance network than was available. This provided the impetus for ANSI to develop the specification for a suitable LAN.

As local area networking matured, different functional areas began to emerge from what used to be a homogeneous network. Each functional area supported a specific task: server connectivity, desktop connectivity, hub interconnectivity, and so forth. Each of these areas has also experienced an increase in the demand for bandwidth. FDDI, with its high data rate and potential for reliability, became a natural choice for connecting servers as well as for interconnecting hubs in the LAN's backbone.

This chapter examines FDDI, its physical media and distance limitations, its frame structures, its mechanics, and some of its benefits. These provide the context for defining FDDI's role in contemporary and future networking.
 
Fundamentals of FDDI
FDDI is an acronym for Fiber Distributed Data Interface, but nobody uses that mouthful of a name. In fact, most people don't even spell out F-D-D-1; they slur the letters together and pronounce it fiddy. FDDI is a robust and reliable LAN technology that dates back to the mid 1980s. FDDI features a 100Mbps data rate and dual counter-rotating rings. These rings can span up to 200 kilometers using fiber-optic cables. Access to this transmission media is regulated through a token-passing scheme that's similar to Token Ring. The token can pass in only one direction.

In the event of a network failure, the repeaters and/or stations are capable of sensing the loss, determining the extent of the network that has lost connectivity, and automatically (but logically) splicing the two rings together. This is known as wraparound or wrapping; this restores connectivity to as much of the network as possible.

FDDI's self-healing capabilities, plus its high data rate, made it the only viable LAN technology for applications with either high bandwidth and/or high reliability requirements. This remained the case for over a decade. Any LAN that needed to support data rates in excess of 16Mbps had to use FDDI. Similarly, any LAN that couldn't afford downtime found FDDI its only viable option. Unfortunately, because it used a fiber-optic transmission media, FDDI was also the most expensive option. This tended to limit its implementation to those highly specialized environments that required either its throughputs or its reliability.

Eventually, other LAN technologies were developed that also reached 10OMbps and beyond. The competition, ATM and Fast Ethernet, were able to meet or exceed FDDI's data rate. This forced FDDI's prices down considerably. Today, FDDI is no longer the elite technology it once was. It remains fairly specialized and is most commonly found in mixed topology LANs. Its two primary uses are to connect servers to multiprotocol switching hubs and even to interconnect switching hubs into a LAN backbone.

Functional Components
FDDI is comprised of four distinct functional components. Each component is defined through its own series of specifications:

These components and their correlation to the OSI Reference Model are presented in Figure below

Media Access Control
As depicted in Figure above, the top layer of FDDI is Media Access Control (MAC). This layer is equivalent to the Data Link Layer of the OSI Reference Model. The MAC sublayer is responsible for defining the media-access methodology and the myriad frame formats. Additionally, the MAC sublayer is also responsible for token and frame generation/management, MAC addressing, and even performing error detection and correction upon receipt of data frames.

Physical Layer Protocol
FDDI's Physical Layer Protocol (PHY) correlates to the upper sublayer functions of the OSI Reference Model's Physical Layer. It's responsible for accepting a data bit stream and converting it to a more suitable format for transmission. This is known as encoding. The actual encoding scheme used is a four-bit/five-bit encoding scheme. This scheme accepts four-bit nibbles (half-octets) from the MAC Layer and encodes each one as a five-bit symbol. This five-bit symbol is what actually gets transmitted. It's important to note that because the MAC Layer is responsible for generating the frames and for framing the data, every piece of the frame is encoded into five-bit symbols.

The PHY is also responsible for setting the clocking for the LAN. FDDI actually runs on a 125MHz transmission clock. The PHY is responsible for generating this transmission clock and synchronizing it across the other stations on the network.

Physical Layer Medium
The Physical Layer Medium (PMD, really!) specifies all the attributes that are expected of the transmission medium: FDDI originally was limited to a single Physical Layer Medium (PMD): 62.5/125-micron diameter multimode fiber-optic cabling. It remained a glass-only technology until the 1990s. Then, the high cost of fiber-optic cable started cutting into its market share. The answer seemed obvious: develop a copper-based PMD that could support the FDDI protocols.

In June of 1990, ANSI formed a working committee to build the specification for a twisted pair PMD (TP-PMD). The TP-PMD was, originally, a proprietary product that grafted the FDDI Layer 2 onto a Category 5 unshielded twisted pair (UTP) Physical Layer. The end result was marketed as CDDI, for Copper Distributed Data Interface. This specification became an ANSI standard in 1994.

A single-mode fiber-optic version (SMF-PMD) has also been developed. Based on 8.3micron diameter fiber-optic cabling and driven by a laser rather than a light-emitting diode (LED), this PMD is more expensive than its multimode counterpart. In its favor is the fact that it can maintain the integrity of the signal for much greater distances: up to 60 kilometers (versus the paltry 2 kilometers of the multimode fiber).

Note: Although the term laser has come into common usage as a noun, it's actually an acronym. It describes the physical process by which the concentrated energy associated with lasers is created. The acronym stands for Light Amplification through Stimulated Emission of Radiation. Laser, howeve much simpler.

Station Management (SMT)
Station Management (SMT, really!) is a separate module that spans the full stack of FDDI protocols. It communicates directly with the MAC, PHY, and PMD layers to monitor and manage the ongoing operation of the station and the ring. The three areas of SMT functionality, as defined in the ANSI X3T9.5 specification, are Together, these three functional areas encompass many different services. They are essential to the normal operation of the station and the FDDI ring. Although there are many others, some of the services provided include the following: Although any given station may have multiple instances of the MAC, PHY, and PMD (as is normal with dual-attached stations), there can be only one SMT.

Building FDDI Networks
FDDI has been stereotyped as having a dual, counter-rotating, ring topology. The truth is, there are several different ways to construct a FDDI network. The dual ring is just one of its many forms. To build more effective FDDI networks, you must understand the various port types and the ways that stations can attach to the network.

Port Types and Attachment Methods
FDDI recognizes four different port types: These port types can be interconnected in a variety of ways. Before you examine them, however, you should understand the different connection types that are supported. The two basic attachment methods that can be used to connect FDDI devices to the network are These attachment methods can be used either with or without repeaters. Attachments can be made between a variety of port configurations. This adds further variety and functionality to the ways that FDDI LANs can be built and used.

Dual-Attached Stations
Dual-attached stations (DASs) feature two sets of media interfaces. This enables a DAS device to have a physical connection to each of FDDI's two rings. Figure below illustrates the way that a dual-attached station connects to the LAN. Each DAS device has two sets of media interface ports, each containing both A and B ports. Each port contains physical connections for two physical media. Therefore, a DAS device actually has four fibers connected to it.


Note: A concentrator is a device that aggregates multiple LAN connections onto a common electrical backplane. The most common type of LAN concentrator is known as a hub. Concentrators, too, can be dual-attached. Consequently, it would be correct to reference both concentrators and stations with the phrase dual-attached(DAS) without specially identifying the devices.

As shown in Figure above the physical device actually becomes an integral part of the two rings, because the network interface card (NIC) provides physical continuity for the two rings between the A and B ports. DAS connections can form a repeaterless, peer-to-peer LAN. This is accomplished by connecting the A port of one device's interface to the B port of another device, and vice versa. The drawback to this is that each DAS device must be powered on and functioning for the rings to be complete. FDDI is capable of wrapping around a break in the ring, but this directly impacts the performance of the entire ring. More significantly, if multiple stations are simultaneously powered down or otherwise out of service, the net result might be two or more smaller ring pairs.

Single-Attached Stations
Single-attached stations (SASs) eliminate the potential performance problems inherent in DAS by getting rid of the wraparound feature. Each SAS device has just a single communications interface, S, with two media ports. The separate fibers are used to transmit and receive. Both fibers terminate at the concentrator, which provides the connectivity to both rings. A single-attached station and its concentrator are illustrated in figure below



Valid Connections
Given this background on port types and attachment methods, a quick survey of the valid types of port connections should complete your understanding of the various connections that can be made with FDDI. Table 8.1 presents all the valid port connection combinations. (See the section "Port Types and Attachment Methods" earlier in this chapter if you can't remember what the letters stand for.)

The only port connection combination that's considered illegal and invalid is M and M. This creates a "ring of trees," which is not very useful. This term is explained in the next section.

Topologies and Implementations
The previously described port types and attachment methods lend themselves to topological and implementation-level variety. Contrary to the persistent myth, FDDI is not just dual, counter-rotating rings. This is, arguably, its most important topology, but there are many other useful topologies and implementations. Some of the more common variations that FDDI networks can use include the following: Each of the first four topologies offers a different combination of performance features and limitations. The fifth one, wraparound, is actually seen only during a network failure.

Dual Ring

The basic dual ring topology, sometimes referred to as a dual ring without trees, is built from dual-attached stations interconnecting directly with each other. This forms a pair of peer-to-peer rings. This is illustrated in Figure below

The disadvantages of relying on each member in the ring for the entire ring's functionality should be obvious. The rings depend equally upon each of the member devices. If any machine is powered down or out of service for any reason, the physical rings are compromised. FDDI detects and automatically limits the extent of the damage, but the point is that risks are inherent in this topology. It should be limited to small, highly specialized environments.

Dual Ring with Trees
The dual ring with trees topology is an enhancement to the dual ring topology. This one features tree-like appendages that grow out of FDDI's dual rings. Developing this topology requires the use of dual-attachment concentrators, single-attachment concentrators, and single-attachment stations. Figure above illustrates a dual ring with trees topology.


The key difference between this topology and the basic dual ring topology is that devices need not connect directly to the rings. Instead, SAS devices connect to single-attached concentrators. These concentrators, in turn, connect to the DAS concentrators that comprise the backbone of the tandem rings.

This topology combines the reliability of dual rings, which automatically wrap around in the event of a failure, with lower costs. SAS components, including concentrators and network interface cards (NICs), are substantially less expensive than their DAS counterparts.

Single Tree
The single tree topology, as its name implies, consists of only a single tree-like grouping of devices. There is no dual ring, nor are there any DAS components. Given that FDDI uses a round-robin token-passing media-access methodology, this tree should be regarded as a logical ring. Tokens still pass in a circular pattern across the network, but the topology is based on a concentrator and, therefore, is star-shaped.

The obvious drawback to this is that there is no redundant path. This directly reduces the reliability of the network. The benefits, however, are many. First, the cost of building a single tree FDDI network is much lower than that of other topologies due to two main factors:

All the devices (concentrators and stations) are the relatively low-cost singleattach variety.

The cost of wiring the LAN's backbone is halved due to the use of only two fibers rather than four.

The other main benefit is reliability. Although this may sound contradictory, given the lack of a second ring, the use of only single-attached devices holds an important implication for the overall reliability of the LAN. The impact of the failure of any given singleattached device is much lower than the failure of an equivalent dual-attached device. If an SAS station fails, the remainder of the network is not affected at all. Similarly, if an SAS concentrator fails, the worst that can happen is that the devices connected to it are isolated from the rest of the network. It doesn't trigger a wraparound. Wraparounds, although touted as a reliability feature, directly impact performance by almost doubling the cable length of the network. Depending on the specific situation, one could make a credible case that auto-recovery via a wraparound is less desirable than simply isolating a few workstations in the event of a concentrator failure.


Dual Home



Dual homing is a specialized use of a dual attachment that provides redundant physical paths for critical networked resources. Such resources may include file and/or application servers, bridges, or even your boss's workstation! Note, however, that dual homing does not necessarily have to include every device on the LAN; therefore, it's not truly a topology. Instead, it's an optional means of implementing LAN connectivity. It can be used very specifically for individual devices rather than broadly for all devices.

This implementation can be used only on a dual ring with trees topology. Each device to be dual homed, by definition, must be DAS capable. Lastly, it must connect to the network via a dual-attached concentrator. Dual homing enables a critical device to have a primary and a less-desirable (from the FDDI protocol's perspective) alternate connection to the LAN.

The Station Management protocols of the dual-homed device activate the primary connection and leave the alternate connection in a standby mode. Each connection terminates at a different dual-attached concentrator. The Station Management protocols can detect this difference in the two connections via their neighbor-discovery mechanisms. Station Management then activates the A-port connection as the primary path and idles the B-port connection. If the A-port connection is lost, for any reason, Station Management attempts to activate the standby connection.

Figure below illustrates a server configured for dual homing within a dual ring with trees topology.



Wraparound
A wraparound isn't really a discrete topology that you would build. Rather, it's automatically constructed by FDDI's station-management mechanisms in the event of either a station failure or a wire path failure. The failure is isolated by logically splicing the primary and alternate rings immediately upstream and downstream from the failure. Implicit in this definition is the fact that only topologies based on a dual ring can wrap around.

Although the mechanics of recovering from either one are similar, there's one fundamental difference between them. Wire path failures can enable all stations to remain active on the wraparound ring network. A failed station, on the other hand, decreases the population of active devices on the network by one.

In Figure below, a wire failure has afflicted Station 2. Its neighboring stations, 1 and 3, wrap their transmissions around this failure onto the secondary ring to preserve the integrity of the loop. The new ring has a total physical media length almost double the size of the original ring. For this reason, the practical maximum media length in a double ring topology should always be half the maximum supportable length for any given media type.



Although Figure above depicts a wire failure in the primary path of a station, similar failures can occur in the LAN's backbone. Dual-attached concentrators, too, can use the alternate ring to wrap their transmissions around wire failures.

If Station 2 in Figure above had failed completely, the wraparound would have looked slightly different. The ring would no longer extend to that device. Instead, it would be wrapped around at Stations I and 3. This is illustrated in Figure below

Devices known as optical bypass switches can be used to prevent station failures from forcing a wraparound. These devices are installed in between the station and the concentrator. In the event of a station failure, these bypass switches maintain the continuity of the wire path without the station.

Network Size
FDDI was designed to be a robust network, capable of supporting high-performance workstations. To maintain high performance for all attached devices, FDDI must impose strict limits on the size of the network. The size of the network can be measured in terms of the following items: All are equally important in developing LANs that can deliver the potential performance of FDDI.

Maximum Devices
The maximum number of devices that any FDDI ring can support is 500. This limitation is actually a function of the maximum allowable propagation delay that FDDI's protocols can endure without compromising its functionality. Each connection adds a measurable amount of propagation delay. The cumulative delays of more than 1,000 physical connections exceed FDDI's delay budget.

Although it may seem simple enough to count 500 devices, the challenge lies in being able to precisely identify a device. In a dual-attached configuration, each device requires two physical connections and, consequently, counts as two connections. Each device in which those connections terminate counts as an additional connection. Therefore, a concentrator port and the device to which it connects actually constitute two devices.

A dual-attached backbone concentrator, one that has no station connections, is counted as two connections. Its ports count as a device only when they are used. A dual-attached station, regardless of whether the two attachments home into the same concentrator or two different concentrators, counts as two devices. Single-attached stations count as only one device.

Ring Length The ANSI X3T9.5 standard does not explicitly state a maximum ring length. Given that the Physical Layer (contrary to popular belief) does not extend to the media itself, this shouldn't be surprising. In keeping with the intent of the Physical Layer, as defined by the OSI Reference Model, the ANSI standard does establish performance parameters that, given any physical media type, impose maximum distances.

In a multimode fiber-optic ring, the total fiber path must be less than 200 kilometers. Unless your FDDI is to span across a large geographic area, like a metropolitan area network (MAN), this limitation shouldn't be much of a constraining factor in your design. What is important to note, however, is the phrase total fiber path.

Two important implications lurk in those key words. First, building a large ring that measures 190 kilometers works until a failure forces a wraparound. Then, the wrapped ring measures something closer to 380 kilometers, and the entire network fails. Therefore, the maximum ring length must always be cut in half when you're designing the LAN.

Second, total fiber path means just that: All the fiber lengths must be added together to determine the total ring length. This includes the main ring and all the branch cables that connect to stations.

Drive Distance
The drive distance is the maximum distance between any two devices. Attenuation is a factor, regardless of which PMD is used. Therefore, the distance between devices must be short enough to guarantee the integrity of the signal upon arrival.

For multimode fiber-optic cabling, the maximum drive distance is 2 kilometers. With single-mode fiber, this increases to 60 kilometers. However, the copper-based PMDs are much more limited. Shielded twisted pair (STP) and Category 5 unshielded twisted pair (UTP) must be kept to less than 100 meters.

e much more limited. Shielded twisted pair (STP) and Category 5 unshielded twisted pair (UTP) must be kept to less than 100 meters.