SHARED FAST ETHERNET

Sales Note - Understanding PDVs and Collision Domains

Why are the configuration rules for 100Mb Ethernet different from 10Mb?

With Ethernet’s Carrier Sense Multiple Access with Collision Detection (CSMA/CD) protocol, LAN devices listen for collisions as they send out packets. If another device transmits before the first device finishes sending, both back off and try again later. For this scheme to work properly, the propagation delay for a packet to travel across the Ethernet segment must be less than the time required to clock the shortest legal packet (64 bytes) onto the LAN. Otherwise, simultaneous senders will not hear each other before they are done sending.

As the clock speed of the signal increases from 10Mb to 100Mb, the time to put a packet onto the LAN decreases. The propagation delay (or packet travel time), however, is determined by the length of the LAN cable . . . including any repeater delays. . . and does not change with the data rate. Therefore, the maximum allowable distance between end stations on a LAN decreases as the speed increases.

What are the limits for 100Mb Ethernet?

At a clock speed of 100Mb/s, it takes 5.12 microseconds to put the shortest legal packet (64 bytes, or 512 bits including headers, etc.) onto the LAN. Therefore, the maximum propagation delay is the same amount of time, 5.12 microseconds or 512 Bit Times. The limit for any shared Ethernet installation is a configuration of cables and repeaters and NICs over which a broadcast packet must travel that does not exceed the maximum propagation delay time. It is convenient to measure the propagation delay in Bit Times (BT). The Fast Ethernet standard is based on this form of measurement.

A collision domain is a cluster of network devices connected together (without bridging) wherein a broadcast Ethernet packet travels to all of the nodes. To be compliant with Fast Ethernet specifications, a collision domain must be less than 512 BT. That is to say, the signal propagation time for the path between any two nodes must be no more than 512 Bit Times. In any real installation, the most significant collision domain path is the longest path in terms of propagation delay since that is the path that is the limiting factor for that installation.

What does "propagation delay time" mean in practical terms?

There is a fixed amount of propagation delay time, 512 BT, that can be used up by the network devices in any path between end points of a collision domain. Each network element in a path uses up some amount of the available time budget for that path. When the total time delay for a path reaches the 512 BT maximum, that path becomes a limiting factor on the network topology since nothing more can be added into that path without violating shared Fast Ethernet specifications.

All network elements introduce some delay into the signal’s path within a collision domain. Each Network Interface Controller (NIC) introduces a delay from where a packet originates, each meter of cable length introduces a delay due to the distance the packet must travel along its way, and each electronic device such as a repeater introduces a delay due to slowing down the packet as it makes its way through the electronics. The delay values are a major characteristic of each network element, and each network element must be measured and rated in terms of its Path Delay Value (PDV).

What are the PDVs for some typical network elements?

NICs are electronic in nature and have PDVs in the same range as other electronic elements. NICs (also called DTEs) that meet the Fast Ethernet specification have a PDV of 50 BT or less. Since there must be two NICs in practically any path through any collision domain . . . one NIC at each end point . . . then 100 BT of the total PDV budget of 512 BT is used up by the NICs. This leaves 412 BT for everything else.

Cables introduce delay that is proportional to their length. Fiber cable is fastest, with a PDV of 1.0 BT per meter. A fiber cable segment that connects two fiber NICs, with no repeater in the path, can be a maximum of 412 meters in length.

Category 5 twisted pair cable introduces slightly more delay than fiber cable since the signal is in the form of electricity traveling over copper rather than light waves traveling through glass. Specifically, CAT 5 cable has a PDV of 1.11 BT per meter. This is 11% more than fiber. (CAT 3 cable has a PDV of 1.14 BT per meter, insignificantly more than CAT 5). Of course, twisted pair cable attenuates the signal as it travels along, and it is limited to 100 meters in length per segment accordingly. A repeater will rejuvenate the signal and boost its strength, enabling a packet to travel another 100 meters over another twisted pair cable segment in the collision domain. Each 100 meters of CAT 5 twisted pair cable has a PDV of 111 BT.

There are typically two twisted pair cables of up to 100 meters each in length in a collision domain path, used to connect the two devices at the end points of the collision domain. These two cable segments use up 222 BT of the PDV budget. So, we are down to 190 BT (512 minus 100 minus 222 = 190) left in the path delay budget for the repeaters and the cables interconnecting the repeaters.

What are the PDVs for Fast Ethernet repeaters?

Since two Class II repeaters must be useable in a collision domain, and since only 190 BT of the delay budget is available for both of them, it can be seen that the PDV of a Class II repeater is 95 (i.e., half of 190) BT. In fact, the Fast Ethernet specification for a Class II repeater is 92 to 95 BT. The specification for a slower repeater that only meets Class I is a PDV of up to 140 BT, and there can practically be only one of them in a 100 Mb/s collision domain. Most Class I and Class II Fast Ethernet repeaters (or hubs, which are the same thing in this case) are designed to meet these specifications, and have very limited installation configurability accordingly.

A better hub design would have a PDV much less than 95 BT. Obviously, a hub with a lower PDV is better because it provides better network configurability. The time delay that is gained from lowering the PDV means that either there can be more total cable length in the longest collision domain path, or there can even be another hub in the path. The configuration limitation of only one (Class I) or two (Class II) hubs is severe, particularly when future growth and expansion of the network must be considered. Hubs with low PDVs that could be configured with three in the PDV budget of 190 BT, i.e., with a PDV of about 60 BT, would permit multi-level cascading and vastly improved network installation limitations.

Are there any Fast Ethernet hubs with low PDVs of 60 to 80 BT? See the Magnum line !!