Tech Arsenal 1

home *** CD-ROM | disk | FTP | other *** search

/ Tech Arsenal 1 / Tech Arsenal (Arsenal Computer).ISO / tek-07 / artisoft.zip / !GD10BS.TXT < prev next >

Wrap

Text File | 1991-12-03 | 15KB | 238 lines

{A Guided Tour of 10BASE-T The recent certification of the 10BASE-T specification churned up a wave of new products and a flood of information -- company position papers and magazine articles explaining what it's all about. While some people see 10BASE-T simply as a newer and more flexible implementation of Ethernet, others are uncertain about what it is and does. And technicians faced with implementing 10BASE-T are justifiably hesitant to proceed without an understanding of the underpinnings of the specification. 10BASE-T gives Ethernet network installers the option to use unshielded twisted-pair (UTP) wiring. It provides a way to install networks using one of the least expensive and most common types of wiring without compromising network performance. The new specification has established UTP as a viable alternative to higher-cost cabling options and provides a touch-stone for interoperability among different vendors' products. Actually, UTP has always been a wiring option. The problem with pre-10BASE-T UTP solutions, however, is that they've all been proprietary. Few of these products provide any level of interoperability with one another. The 10BASE-T specification resolves these incompatibilities, specifying how devices that communicate on a 10BASE-T segment are to perform. 10BASE-T is actually a supplement to the standard 802.3 specification, which established the requirements for both thin and thick coax networks. This means that the underpinnings of 802.3 - the data transmission clock rate, for example -- remain unchanged while the ability to operate with UTP wiring has been added. To support reliable transmission of Ethernet signaling over UTP wiring, four problems needed to be resolved: electromagnetic emissions (from transmitting devices), susceptibility (to other devices transmitting in the same frequencies as the 10BASE-T signal), crosstalk (between the UTP wires) and jitter (explained later). The 10BASE-T committee was able to overcome these problems by leveraging off the experience of vendors who had already provided UTP networking solutions. However, having a specification alone didn't necessarily ensure that the products from all vendors providing 10BASE-T products would work together. Like most IEEE specifications, 10BASE-T does not actually tell a vendor how to design compliant devices; instead, the specification simply describes how such a device should operate. {10BASE-T Defined 10BASE-T defines two basic network components: the wiring and the devices that terminate the ends of a wiring segment, known as media attachment units (MAUs). 10BASE-T cable segments use four unshielded wires, typically at 22 to 26 AWG (between 0.6 and 0.4mm thick). These wires are attached to lines 1, 2, 3, and 6 of an ISO-compliant physical interface -- i.e., an RJ-45 jack. These lines correspond to "tip and ring" receive and transmit, respectively ("tip and ring" are telephone terms for essentially plus and minus). Wiring lengths run from zero distance -- two RJ-45 jacks wired together -- to some theoretical limit. 10BASE-T provides a list of specifications for a cable segment, concluding with the statement that these specifications "are generally met by 100 meters of 0.5mm telephone twisted pair." It's theoretically possible for a higher grade of wiring under ideal conditions to support greater segment lengths. For example, AT&T's network interface cards and hubs can drive segments up to 150 meters. With special AT&T cabling that distance can be extended to 200 meters. The downside of such segment-extension schemes is that they are proprietary, operating solely with a single vendor's equipment, which is exactly what 10BASE-T was designed to avoid. If you're not using a scheme like AT&T's, however, you'll probably need to make a careful analysis of your existing UTP wire before using it for 10BASE-T. Another misunderstanding about 10BASE-T is its supposed susceptibility to electromagnetic emissions. A 10BASE-T signal transmits in a frequency range of 10 to 200 Mhz. In a common office environment, few devices emit electromagnetic radiation in that range with enough power to affect a signal on a 10BASE-T UTP segment. While it's theoretically possible that an FM radio transmitter's 80 Mhz to 108 Mhz broadcast could affect 10BASE-T transmissions, your LAN would have to be within touching range of the radio transmitter to be affected. {Establishing Connections Included in the 10BASE-T specification are the required capabilities for MAUs. Any device intended to transmit over a 10BASE-T segment must interface to the segment as a MAU. Typical 10BASE-T devices that act as MAUs include network interface cards (NICs) and hubs. the 10BASE-T specification also has a name for devices that can initiate the transmission of information over cable segments: DTEs, for Data Terminating Equipment. A typical DTE is a LAN PC with a NIC. Physically, 10BASE-T NICs look almost exactly like the 10BASE-2 and 10BASE-5 NICs you've used previously, because the manufacture of a 10BASE-T NIC doesn't represent any unusual redesign or production changes. Standard Ethernet NICs usually provide both BNC and AUI interfaces for 10BASE-2 and 10BASE-5 connections, respectively. Most 10BASE-T vendors have simply removed the NICs BNC connector and a few chips, replacing these with an MAU chipset and RJ-45 connector. The network hub is what actually supports the interoperability between different DTE's. Different vendors call the hub a multiport repeater, concentrator or wiring center. 10BASE-T hubs are active devices; that is they interact with the signal on a 10BASE-T segment, performing vital functions for packet retransmissions. In this role, the hub serves as a signal repeater. To work as a repeater port, the MAUs phase-locked loop (PLL) circuitry senses a series of changes in voltage on the 10BASE-T wire. This set of signals, called the preamble, tells the MAU that a packet is to follow. The PLL starts its clock, knowing that the timing for the packet signals will match the preamble timing. The MAU also knows the signaling requirements for the packet header, as well as the maximum length a packet can be. If the timing for the preamble signals and the packet signals are out of phase, a condition known as "jitter" exists. The hub is expected to remove any jitter, regenerate the preamble and clock, amplify signal symmetry and send the packet on its way. The hub must also serve as an active filter, rejecting packets that are severely distorted. The 10BASE-T specification mentions a "jitter-budget", which defines the jitter limits for various components on each 10BASE-T segment. If the aggregate amount of jitter generated by all these components exceeds the budget, a device may not be able to transmit on that segment. How do MAUs know when they can transmit? 10BASE-T MAUs use a state machine architecture. They determine whether the 10BASE-T segment to which they are connected is viable for communications, and they know whether there is another functioning MAU at the other end of the segment. During traffic idle states, a MAU sends out a "link integrity" pulse to ensure that there's another active MAU at the other end of the link. External conditions can cause a MAU to determine that it is unable to transmit under the required conditions. For example, if an active telephone cable carrying a phone line voltage is inadvertently connected to a 10BASE-T segment, the MAUs at either end of the segment should shift in to a "link fail" state because they "know" they can't transmit on such a line. If a DTE gets stuck in transmit for a period longer than 150 ms (known as "jabber"), a MAU is required to disable transmit and loop back, shift status to "link bad" wait for a period of 0.25 to 0.75 seconds, then re-enable the link. Most vendors put link integrity LEDs on the backplane connector of their NICs and above the connector ports for their hubs to allow diagnosis of problem situations. Moving to a hub-based wiring scheme has a number of advantages, the greatest of which is the potential for network management. A variety of vendors offer different schemes for gathering network transmission statistics, monitoring problems and isolating faults. Some of these solutions are proprietary, while others use more widely adopted protocols such as SNMP. {Installation If you've installed Ethernet NICs before, you'll find little difference in the installation of NICs for 10BASE-T. As a matter of fact, aside from the different connectors, you may have a hard time telling the 10BASE-T boards apart from their 10BASE-2 or 10BASE-5 counterparts. Note that you don't have to discard your investment in 10BASE-2 and 10BASE-5 Ethernet NICs to implement 10BASE-T. Interface devices known as "micro-MAUs" can be attached directly to the AUI port on most Ethernet NICs that allow a 10BASE-T connection through the existing NIC. You will also find that the cost for wiring new connections will probably be lower. The wire itself is less expensive and more plentiful. Vendors for the installation of telephone-type wiring abound and you'll no longer have to explain patiently to the installer the differences among grades of coax cable. When it comes to hooking the wiring together, however, you may need to do some adapting. Unless you've used multiport repeater configurations for your 10BASE-2 or 10BASE-5 Ethernet network before, you'll find that the requirements of star-wiring include a significant mental shift. You have to redesign your network to support a hub-based layout, pull workgroup connections to a wiring closet, then run cables between closets. In addition, you'll have to take into account the cost of the hubs themselves. Also, every time you need to add a new hub, your per-station wiring costs goes up. You'll also increase your costs as you add newer capabilities to your hubs, including network management and bridging to other topologies and locations. To help decide whether to use 10BASE-T, answer the following questions: {* How long are your existing wiring segments?} In most cases, you won't have to worry: AT&T reports that 95 percent of all phone connections have a telephone well within 100 meters of a wiring closet. {* What's the quality of your wiring?} Not all UTP wire is created equal and it will affect the maximum length of cable segments. {* Are you sure you'll be using RJ-45 jacks?} Even though 10BASE-T needs only four wires,don't make the mistake of using RJ-11 jacks. An RJ-11 connector will fit into an RJ-45 slot, but the wires won't correspond. {* Does your wiring run through many series of punchdown blocks?} If so, make sure the blocks have been wired correctly. In many buildings with excess telephone wiring, the connections have never been used. Be careful to test the wiring you intend to use for 10BASE-T to make sure that no active telephone lines are connected. If line segments carry tinging or battery voltages, devices using them will be unable to communicate. Document everything. When designing your network, remember that 10BASE-T is still a CSMA/CD system. The hubs are not bridges; they don't filter traffic, except at a very low level. And you'll still have all the attendant issues to deal with that you do on 10BASE-2 and 10BASE-5 networks, including the potential for waves of collisions, packet fragmentation and broadcast storms. If you've already installed a UTP network that uses a pre-10BASE-T scheme (StarLAN 10, for example), get your vendor to tell you how to link to a compliant system. In some cases you'll find that your existing wiring can't meet the standards imposed by 10BASE-T. Ideally, you'll end up using 10BASE-T in situations where you have control over the new wiring of a workgroup, say in a "cubicle jungle" where new phone and other wiring must be run anyway. This will allow you to control the quality and segment lengths of your UTP wiring, so you can design your 10BASE-T network from scratch.