Network Administrator Network Security

The 1000-Mbps Ethernet or Gigabit Ethernet standards represent transmission using both fiber and copper media. The 1000BASE-X standard, IEEE 802.3z, specifies 1 Gbps full duplex over optical fiber. The 1000BASE-X standard, IEEE 802.3z, specifies 1 Gbps full duplex over optical fiber.

1000BASE-TX, 1000BASE-SX, and 1000BASE-LX use the same timing parameters, as shown in Figure . They use a 1 nanosecond (0.000000001 seconds) or 1 billionth of a second bit time. The Gigabit Ethernet frame has the same format as is used for 10 and 100-Mbps Ethernet. Depending on the implementation, Gigabit Ethernet may use different processes to convert frames to bits on the cable. Figure shows the Ethernet frame formats.

The differences between standard Ethernet, Fast Ethernet and Gigabit Ethernet occur at the physical layer. Due to the increased speeds of these newer standards, the shorter duration bit times require special considerations. Since the bits are introduced on the medium for a shorter duration and more often, timing is critical. This high-speed transmission requires frequencies closer to copper medium bandwidth limitations. This causes the bits to be more susceptible to noise on copper media.

These issues require Gigabit Ethernet to use two separate encoding steps. Data transmission is made more efficient by using codes to represent the binary bit stream. The encoded data provides synchronization, efficient usage of bandwidth, and improved Signal-to-Noise Ratio characteristics.

At the physical layer, the bit patterns from the MAC layer are converted into symbols. The symbols may also be control information such as start frame, end frame, medium idle conditions. The frame is coded into control symbols and data symbols to increase in network throughput.

Fiber-based Gigabit Ethernet (1000BASE-X) uses 8B/10B encoding which is similar to the 4B/5B concept. This is followed by the simple Non-Return to Zero (NRZ) line encoding of light on optical fiber. This simpler encoding process is possible because the fiber medium can carry higher bandwidth signals.

Because Cat 5e cable can reliably carry up to 125 Mbps of traffic, getting 1000 Mbps (Gigabit) of bandwidth was a design challenge. The first step to accomplish 1000BASE-T is to use all four pairs of wires instead of the traditional two pairs of wires used by 10BASE-T and 100BASE-TX. This is done using complex circuitry to allow full duplex transmissions on the same wire pair. This provides 250 Mbps per pair. With all four-wire pairs, this provides the desired 1000 Mbps. Since the information travels simultaneously across the four paths, the circuitry has to divide frames at the transmitter and reassemble them at the receiver.

The 1000BASE-T encoding with 4D-PAM5 line encoding is used on Cat 5e or better UTP. Achieving the 1 Gbps rate requires use of all four pairs in full duplex simultaneously. That is the transmission and reception of data happens in both directions on the same wire at the same time. As might be expected, this results in a permanent collision on the wire pairs. These collisions result in complex voltage patterns. With the complex integrated circuits using techniques such as echo cancellation, Layer 1 Forward Error Correction (FEC), and prudent selection of voltage levels, the system achieves the 1Gigabit throughput.

In idle periods there are nine voltage levels found on the cable, and during data transmission periods there are 17 voltage levels found on the cable. With this large number of states and the effects of noise, the signal on the wire looks more analog than digital. Like analog, the system is more susceptible to noise due to cable and termination problems.

The data from the sending station is carefully divided into four parallel streams, encoded, transmitted and detected in parallel, and then reassembled into one received bit stream. Figure represents of the simultaneous full duplex on four-wire pairs. 1000BASE-T supports both half-duplex as well as full-duplex operation. The use of full-duplex 1000BASE-T is widespread.

The distance limitations of full-duplex links are only limited by the medium, and not the round-trip delay. Since most Gigabit Ethernet is switched, the values in Figures and are the practical limits between devices. Daisy-chaining, star, and extended star topologies are all allowed. The issue then becomes one of logical topology and data flow, not timing or distance limitations.

A 1000BASE-T UTP cable is the same as 10BASE-T and 100BASE-TX cable, except that link performance must meet the higher quality Category 5e or ISO Class D (2000) requirements.

Modification of the architecture rules is strongly discouraged for 1000BASE-T. At 100 meters, 1000BASE-T is operating close to the edge of the ability of the hardware to recover the transmitted signal. Any cabling problems or environmental noise could render an otherwise compliant cable inoperable even at distances that are within the specification.

It is recommended that all links between a station and a hub or switch be configured for Auto-Negotiation to permit the highest common performance. This will avoid accidental misconfiguration of the other required parameters for proper Gigabit Ethernet operation.

With the frame format and other Ethernet Layer 2 specifications compatible with previous standards, 10GbE can provide increased bandwidth needs that are interoperable with existing network infrastructure.

A major conceptual change for Ethernet is emerging with 10GbE. Ethernet is traditionally thought of as a LAN technology, but 10GbE physical layer standards allow both an extension in distance to 40 km over single-mode fiber and compatibility with synchronous optical network (SONET) and synchronous digital hierarchy (SDH) networks. Operation at 40 km distance makes 10GbE a viable MAN technology. Compatibility with SONET/SDH networks operating up to OC-192 speeds (9.584640 Gbps) make 10GbE a viable WAN technology. 10GbE may also compete with ATM for certain applications.

To summarize, how does 10GbE compare to other varieties of Ethernet?

• Frame format is the same, allowing interoperability between all varieties of legacy, fast, gigabit, and 10 Gigabit, with no reframing or protocol conversions.
• Bit time is now 0.1 nanoseconds. All other time variables scale accordingly.
• Since only full-duplex fiber connections are used, CSMA/CD is not necessary
• The IEEE 802.3 sublayers within OSI Layers 1 and 2 are mostly preserved, with a few additions to accommodate 40 km fiber links and interoperability with SONET/SDH technologies.
• Flexible, efficient, reliable, relatively low cost end-to-end Ethernet networks become possible.
• TCP/IP can run over LANs, MANs, and WANs with one Layer 2 Transport method.

The basic standard governing CSMA/CD is IEEE 802.3. An IEEE 802.3 supplement, entitled 802.3ae, governs the 10GbE family. As is typical for new technologies, a variety of implementations are being considered, including:

• 10GBASE-SR – Intended for short distances over already-installed multimode fiber, supports a range between 26 m to 82 m
• 10GBASE-LX4 – Uses wavelength division multiplexing (WDM), supports 240 m to 300 m over already-installed multimode fiber and 10 km over single-mode fiber
• 10GBASE-LR and 10GBASE-ER – Support 10 km and 40 km over single-mode fiber
• 10GBASE-SW, 10GBASE-LW, and 10GBASE-EW – Known collectively as 10GBASE-W are intended to work with OC-192 synchronous transport module (STM) SONET/SDH WAN equipment.

The IEEE 802.3ae Task force and the 10-Gigabit Ethernet Alliance (10 GEA) are working to standardize these emerging technologies.

10-Gbps Ethernet (IEEE 802.3ae) was standardized in June 2002. It is a full-duplex protocol that uses only optic fiber as a transmission medium. The maximum transmission distances depend on the type of fiber being used. When using single-mode fiber as the transmission medium, the maximum transmission distance is 40 kilometers (25 miles). Some discussions between IEEE members have begun that suggest the possibility of standards for 40, 80, and even 100-Gbps Ethernet.

In response to these issues of synchronization, bandwidth, and Signal-to-Noise Ratio, 10-Gigabit Ethernet uses two separate encoding steps. By using codes to represent the user data, transmission is made more efficient. The encoded data provides synchronization, efficient usage of bandwidth, and improved Signal-to-Noise Ratio characteristics.

Complex serial bit streams are used for all versions of 10GbE except for 10GBASE-LX4, which uses Wide Wavelength Division Multiplex (WWDM) to multiplex four bit simultaneous bit streams as four wavelengths of light launched into the fiber at one time.

Figure represents the particular case of using four slightly different wavelength, laser sources. Upon receipt from the medium, the optical signal stream is demultiplexed into four separate optical signal streams. The four optical signal streams are then converted back into four electronic bit streams as they travel in approximately the reverse process back up through the sublayers to the MAC layer.

Currently, most 10GbE products are in the form of modules, or line cards, for addition to high-end switches and routers. As the 10GbE technologies evolve, an increasing diversity of signaling components can be expected. As optical technologies evolve, improved transmitters and receivers will be incorporated into these products, taking further advantage of modularity. All 10GbE varieties use optical fiber media. Fiber types include 10µ single-mode Fiber, and 50µ and 62.5µ multimode fibers. A range of fiber attenuation and dispersion characteristics is supported, but they limit operating distances.

Even though support is limited to fiber optic media, some of the maximum cable lengths are surprisingly short. No repeater is defined for 10-Gigabit Ethernet since half duplex is explicitly not supported.

As with 10 Mbps, 100 Mbps and 1000 Mbps versions, it is possible to modify some of the architecture rules slightly. Possible architecture adjustments are related to signal loss and distortion along the medium. Due to dispersion of the signal and other issues the light pulse becomes undecipherable beyond certain distances.

10BASE5 uses Manchester encoding. It has a solid central conductor. Each of the maximum five segments of thick coax may be up to 500 m (1640.4 ft) in length. The cable is large, heavy, and difficult to install. However, the distance limitations were favorable and this prolonged its use in certain applications.

Because the medium is a single coaxial cable, only one station can transmit at a time or else a collision will occur. Therefore, 10BASE5 only runs in half-duplex resulting in a maximum of 10 Mbps of data transfer.

Figure illustrates one possible configuration for a maximum end-to-end collision domain. Between any two distant stations only three repeated segments are permitted to have stations connected to them, with the other two repeated segments used only as link segments to extend the network.

10BASE2 also uses Manchester encoding. Computers on the LAN were linked together by an unbroken series of coaxial cable lengths. These lengths were attached by BNC connectors to a T-shaped connector on the NIC.

10BASE2 has a stranded central conductor. Each of the maximum five segments of thin coax may be up to 185 meters long and each station is connected directly to the BNC “T” connector on the coax.

Only one station can transmit at a time or else a collision will occur. 10BASE2 also uses half-duplex. The maximum transmission rate of 10BASE2 is 10 Mbps.

There may be up to 30 stations on any individual 10BASE2 segment. Out of the five consecutive segments in series between any two distant stations, only three may have stations attached

Although hubs may be linked, it is best to avoid this arrangement. This is to prevent exceeding the limit for maximum delay between distant stations. When multiple hubs are required, it is best to arrange them in hierarchical order as to create a tree structure. Performance will be improved if fewer repeaters separate stations.

An architectural example is shown in Figure . All distances between stations are acceptable. However, the total distance from one end of the network to the other, places the architecture at its limit. The most important aspect to consider is how to keep the delay between distant stations to a minimum, regardless of the architecture and media types involved. A shorter maximum delay will provide better overall performance.

10BASE-T links can have unrepeated distances up to 100 m. While this may seem like a long distance, it is typically “used up” when wiring an actual building. Hubs can solve the distance issue but will allow collisions to propagate. The widespread introduction of switches has made the distance limitation less important. As long as workstations are located within 100 m of a switch, the 100 m distance starts over at the switch.

The original coaxial Ethernet used half-duplex transmission so only one device could transmit at a time. However, in 1997, Ethernet was expanded to include a full duplex capability that allowed more than one PC on a network to transmit at the same time. Switches increasingly replaced hubs. These switches had the capability of full duplex and rapid handling of Ethernet frames.

100BASE-TX uses 4B/5B encoding, which is then scrambled and converted to multi-level transmit-3 levels or MLT-3. In the example, the highlighted window shows four waveform examples. The top waveform has no transition in the center of the timing window. No transition indicates that a binary 0 is present. The second waveform shows a transition in the center of the timing window. A binary 1 is represented by a transition. The third waveform shows an alternating binary sequence. The absence of binary transition indicates a binary 0, and the presence of a transition indicates a binary 1. Rising or falling edges indicate 1s. Very steep signal changes indicate 1s. Any noticeable horizontal line in the signal indicates a 0.

Figure shows the pinout for a 100BASE-TX connection. Notice that the two separate transmit-receive paths exist. This is identical to the 10BASE-T configuration.

100BASE-TX carries 100 Mbps of traffic in half-duplex mode. In full-duplex mode, 100BASE-TX can exchange 200 Mbps of traffic. The concept of full duplex will become increasingly important as Ethernet speeds increase.

The timing, frame format, and transmission are all common to both versions of 100 Mbps Fast Ethernet. 100BASE-FX also uses 4B/5B encoding. In Figure notice the highlighted waveform in the example. The top waveform has no transition, which indicates that a binary 0 is present. In the second waveform, a transition is in the center of the timing window. A binary 1 is represented by a transition. In the third waveform, there is an alternating binary sequence. In this example it is more obvious that no transition indicates a binary 0 and the presence of a transition is a binary 1.

Figure summarizes a 100BASE-FX link and pinouts. Fiber pair with either ST or SC connectors is most commonly used.

200 Mbps transmission is possible because of the separate Transmit and Receive paths in 100BASE-FX optical fiber.

A Class I repeater may introduce up to 140 bit-times of latency. Any repeater that changes between one Ethernet implementation and another is a Class I repeater. A Class II repeater may only introduce a maximum of 92 bit-times latency. Because of the reduced latency it is possible to have two Class II repeaters in series, but only if the cable between them is very short.

As with 10 Mbps versions, it is possible to modify some of the architecture rules for 100 Mbps versions. However there is virtually no allowance for additional delay. Modification of the architecture rules is strongly discouraged for 100BASE-TX. 100BASE-TX cable between Class II repeaters may not exceed 5 meters. Links operating in half duplex are not uncommon to find in Fast Ethernet. However, half duplex is undesirable because the signaling scheme is inherently full duplex.

Figure shows architecture configuration cable distances. 100BASE-TX links can have unrepeated distances up to 100 m. The widespread introduction of switches has made this distance limitation less important. If workstations are located within 100 m of a switch, the 100 m distance starts over at the switch. Since most Fast Ethernet is switched, these are the practical limits between devices.

Cisco Systems, Inc.