Local area network cables. Radio channel, infrared channel, microwave channel. Coaxial network cable
Maybe someone will consider this material untimely; indeed, while “the entire civilized world” is switching to Gigabit Ethernet, we are suddenly releasing material dedicated to 100-megabit twisted-pair networks. However, let's not rush to conclusions. The civilized world is, of course, good, but if you look at the LAN in the computerized office of an “average” domestic company, you immediately understand one thing: “Learning is light, and the unlearned are…”.
Every specialist responsible for a local network (or, in a particular case, for creating it “from scratch”) repeatedly has to answer a difficult question: can it cope with the tasks assigned to it? Will it be up to the new tasks we might someday want to assign to it? How to insure yourself against the need for expensive network modifications for at least a few years? How to ensure the possibility of its modernization with “little loss”? When everything works like a clock, the work of a network administrator as an overseer and regulator of traffic between users is not burdensome and quite simple. But when problems arise, it is he who often finds himself sitting on hot coals...
In this material, we tried to take the position of a person who has an idea of what “computer hardware” is, but who understands networks, to put it mildly, superficially. After all, not everyone network administrator begins its activities after graduating from the relevant faculty of the university, passing certification courses and subsequent six-month internship under the supervision of “senior comrades, smart and sensitive.” In our country, alas, the most popular IT profession is still the “computer specialist”: “Yes, we have a programmer... Yes, he also changes the cartridges in the printer... Yes, he will install the OS and software if necessary. What are you saying? Not a "programmer"? You know, to tell the truth, that’s what I call them all...” And when the number of computers in the office becomes more than three, it is precisely these “young specialists” (how conveniently the term from Soviet times came here!) that the company’s management often sets the task: “Make a network. Fast. Cheap. And reliably! And they find themselves in the position of a kitten, caught not only in a pool, but in the very middle of a whirlpool... LAN: what is this?
To begin with, it is useful to familiarize yourself with the “canonical” definition. So, a local area network is a distributed system built on the basis of a local communication network and designed to ensure physical connectivity of all system components located at a distance not exceeding the maximum for this technology. In essence, a LAN implements the technology of integration and collective use of computing resources. The main advantages of such distributed systems are the following: high data processing performance, increased modularity and expandability, reliability, survivability, constant availability and low cost. Also, such a definition cannot be considered complete without focusing on ease of reconfiguration and minimizing the costs of further modernization.
"On the Top"
In reality, a typical “average small LAN” consists of three conventional classes of devices:
- computers with network adapters installed in them;
- “cable management”, to which we include the network cables themselves, patches, patch panels and (optional) cabinets or racks;
- active network equipment, which can also be placed in cabinets or racks, including the same ones as patch panels (usually switches and/or hubs).
Again, in the simplest case, all computers on the network are simply connected to one hub or switch (directly or through a patch panel - we are not interested in this yet). In a more complex case, several hubs or switches are connected to each other via an Uplink connector (so-called “cascading”). In an even more complex scenario, several hubs (switches) form network segments, “brought together” by another dedicated switch (and here you don’t have to add “or a hub”; a competent network administrator, as a rule, in this capacity avoids using them). This is where we will finish the list of the simplest and most common options for building a LAN for now.
By the way, it seems appropriate to remind network specialists that in this material we have to make many simplifications due to its focus on the widest range of readers. Of course, following the canons and clarity of definitions is not bad, but I still don’t want to put a potential novice network administrator in the position of the hero Mark Twain, who once said: “Until they explained to me in geometry class that a circle is a collection points located at the same distance from the center, I knew well what a circle was!
Network "on the knee"
At the dawn of the “network era,” deviations from cable network standards were often allowed when building domestic LANs. Often the reason for this was poverty (the fiber optic cable system and equipment, although significantly cheaper, were not equal in cost to “copper” solutions), sometimes negligence, and in most cases, basic technical illiteracy. And if sometimes you have to put up with the first reason (lack of money), then the next two are quite possible to eliminate, since they are caused exclusively by the “human factor”.
However, oddly enough, networks built in violation of standards worked for the time being! However, only for the time being. For example, we have not yet had to replace any network device (network adapter, hub, etc.). And here, after the replacement, the entire network suddenly began to “fever” in an unpredictable way... At the same time, it could work normally with all applications except one, and the administrator’s attempt to “press it to the wall” was worth both time and, especially, nerves. But it was not the application or the network card that was to blame, but the entire network. Or rather, those who selected the equipment, installed the cable and put the system into operation without thinking (or not suspecting?) about the standards. Even more serious problems arose when trying to transfer a network built “with deviations” from Ethernet to Fast Ethernet. Indeed, at high speeds, the LAN becomes much more demanding on the quality of the cable system, and those assumptions that were “goodbye” at 10 Mbps often plunge a 100-megabit network into a state of stupor.
But what if it’s “wise”?
Thus, first of all, it is worth remembering once and for all that the design and installation of any LAN implies, first of all, strict adherence to the relevant standards and recommendations, which ensures its normal functioning not in “some”, but in all cases provided for by these standards.
- Modern wired LANs are implemented on the basis of twisted pairs and fiber optic cables.
- Topology defines the general structure of relationships between elements and characterizes the complexity of the interface.
- Methods of access to the physical medium are divided into random and deterministic and depend on the network topology.
First, a little history. It so happens that to organize the interaction of nodes in local networks built on the basis of classical technologies (Ethernet, Token Ring, FDDI), developed 15–20 years ago, communication channels shared between a group of computers (common bus, ring), access to which are provided using a special algorithm (usually a random access method or a method with transmission of an access token over a ring), i.e. based on the principle of using shared media or supporting it.
Against, modern standards and local network technologies insist on partial or complete abandonment of the use of shared data transmission media and the transition to the use of individual computer communication channels with communication devices networks. That is, the same way as is done in the telephone networks we are accustomed to, where each telephone set is connected to a switch on the PBX by an individual line. Application-oriented technologies individual lines communications are Fast- and Gigabit Ethernet, 100VG-AnyLAN, ATM and switching modifications of the already mentioned classic technologies. Note that some of them, for example l00VG-AnyLAN, remained in the minds of domestic “network builders” as nothing more than sounding exotic.
Fast Ethernet as a development of classic Ethernet
Basics of the currently most popular technology for building local computer networks Ethernet were developed by Xerox Corporation's Palo Alto Research Center (PARC) in the mid-1970s. Its specifications were prepared for industrial implementation by members of the DIX consortium (DEC, Intel, Xerox) and adopted as the basis for the development of the IEEE 802.3 standard in 1980. Pay attention to the dates! In fact, we can state that not much has changed since those times...
10 Mbit Ethernet has been satisfactory for most users for about 15 years. However, in the early 90s, its insufficient bandwidth began to be felt, and the next significant step in the development of classical Ethernet technology was Fast Ethernet. In 1992, a group of network equipment manufacturers, including such leaders as SynOptics, 3Com and several others, formed the Fast Ethernet Alliance to develop a standard for a new technology that would summarize and generalize the achievements of individual companies in the field of Ethernet-compatible high-speed standard. At the same time, work began at the IEEE Institute to standardize the new technology. After breaking a lot of copies, in May 1995, the IEEE committee adopted the Fast Ethernet specification as the 802.3u standard (adding chapters 21 to 30 to the 802.3 base document). This played a decisive role in the future fate of the technology, as it ensured the continuity and consistency of the 10Base-T and 100Base-T networks.
10- to 100Base-T
Differences at the physical and data link layers of the OSI model protocol stack
From the figure (in terms and categories of the seven-layer OSI model) it is clear that the differences between Fast Ethernet and Ethernet are concentrated on the physical layer. The 100Base-T (802.3u) standard established three different specifications for the physical layer to support the following types cable systems:
- 100Base-TX for two-pair UTP Cat. 5 or shielded twisted pair STP Type 1;
- 100Base-T4 for four-pair UTP Cat. 3, 4 or 5;
- 100Base-FX for multimode fiber optic cable.
Physical interfaces of the Fast Ethernet IEEE 802.3u standard and their main characteristics
* OmV single-mode fiber, MmV multimode fiber.
**Distance can only be achieved with full duplex communication.
*** It is not widespread in our country due to the fundamental impossibility of supporting a duplex transmission mode.
Full duplex mode
New in this standard (for network nodes that support the FX and TX specifications) is also the recommendation to provide full-duplex mode when connecting a network adapter to a switch or when connecting switches directly to each other. The specificity of the operation is that each node simultaneously transmits and receives data frames via Tx and Rx channels. Transfer speed up to 200 Mbps. Today, many manufacturers declare the release of both network adapters and switches that support this mode. However, alas, due to different understandings of the mechanisms of its implementation, in particular how to manage the flow of personnel, these products do not always work correctly with each other. By the way, for those who are accustomed to reading articles “diagonally”: pay attention to the method of connecting which devices become possible work network cards in full duplex mode. Hint: hubs are not included in this list. And for good reason.
Hubs and switches
The Fast Ethernet network that is “closest” to us, built on the basis of a hub (in the jargon of networkers “hub”, from the English hub) and uniting several dozen users, often turns out to be “ineffective” in the sense that the data transfer speed in it will be unacceptably low , and some clients may be denied access to network resources altogether. This is due to an increase in the number of collisions (see glossary) and an increase in access latency. After all, a hub is a regular amplifier (receiver-repeater) of an electrical signal; sometimes even manufacturers, in the old fashioned way, label it as “(Fast) Ethernet repeater.” Having received a network packet from one port (i.e. from a computer that is connected to a given port), it broadcasts it to all other ports simultaneously (the principle can be roughly defined as “I transmitted it to everyone, which means that the one who needs it will also reach ").
Switch (also known as “switch” in common parlance, from the English switch) more intelligent device: it has its own processor, internal high-performance bus and buffer memory. While a hub simply forwards packets from one port to all others, a switch performs targeted forwarding of packets between two ports based on the destination MAC address. This allows you to increase network performance, as it minimizes the possibility of collisions, allows you to handle packet forwarding between several ports simultaneously, etc.
Having noticed that recently the cost of switches for Fast Ethernet networks is gradually approaching the cost of hubs from the beginning of last year, let us briefly summarize the advantages of networks built using them:
- Network performance increases by dividing it into addressable (logically) interconnected segments.
- The possibility of interception of passwords and other transmitted/received information by a third party is excluded (remember that when using a hub, any packet is broadcast to all computers connected to it).
If it is possible to name any reason (other than the conservatism of the network owner) limiting the widespread use of switches, it is still their higher cost than that of hubs. Although in fairness it is worth noting that soon it seems that we will have no choice: an increasing number of network equipment manufacturers are simply abandoning hubs, preferring to release new, cheaper switch models or reduce prices on those already produced.
Gigabit at the end of the tunnel?
Of course, it's 2002, and even in our country, more and more corporate customers are already seriously looking at Gigabit Ethernet as the base standard for their networks. But still, in terms of mass popularity, it is Fast Ethernet technology (the subject of our attention today) that continues to hold its leading position. Moreover, domestic experts predict a long life even for “old” Ethernet networks (10 Mbps), predicting their gradual modernization to 100 Mbps “big brother”, the speed capabilities of which a typical office network will probably be completely satisfied with for many more years. Of course, unless you plan to hold teleconferences with dozens of participants. However, in this regard, in the process of preparing the material, we even came up with one technical “joke”: the cost of equipment that will allow you to load a network based on Gigabit Ethernet with work often even exceeds the cost of deploying this very network. In addition, it is worth noting that designing, installing and deploying a Gigabit Ethernet network is hardly where you need to start “practical experiments in arranging a LAN.”
From the history of Ethernet (for those interested)
Few people know that the emergence of Ethernet is inextricably linked with such cornerstones of the modern computer industry as Fabless and Core Logic. These two concepts are difficult to translate into Russian while maintaining the laconicism of the English language.
At a time when there was a misconception that controller design (essentially Core Logic) was the domain of the semiconductor industry, not without the help of the hero of our story, Gordon A. Campbell, the idea of independent development, located at the facilities of third-party manufacturers, materialized. Since then, “horselessness” (read Fabless) in the computer world is not considered a sin, but is revered as a property of a sharp mind.
For mutual understanding of developers and manufacturers, with the blessing of Gordon Campbell, a language for describing the internal structure of a chip, VHDL (Very High Definition Language), arose and developed. And the very concept of a chip rightfully occupies an honorable place in the almost endless list of Mr. Campbell’s ingenious initiatives.
In addition to the above, the merits of Gordon Campbell in a brief summary look like this:
- the idea of reprogrammable controllers such as EEPROM;
- idea and implementation of PC-on-chip;
- organizational work on the formation of Palm Corp.;
- development of the first IBM-compatible video controller;
- fundamental works in the field of 3D graphics;
- participation in the founding of 3Dfx Interactive.
The time has come to name the company “involved” in the successes of Mr. Campbell , by the way, and organized by him: Chips & Technologies Inc. In close collaboration with Novell, more than ten years ago, a product was born that would long define the structure of modern networking technologies, Novell Eagle. Today, the abbreviation NE2000 is known to everyone involved in network technologies.
Novell developed software model driver support for Ethernet, and Chips & Technologies took on the programming of semiconductor logic. Manufacturing was entrusted to National Semiconductor. This is how a chipset appeared, consisting of three components:
- DP8990 (Network Interface Controller, NIC) interface for connecting to the local bus of a personal computer;
- DP8991 (Serial Network Interface, SNI) data serialization using Manchester encoding and collision servicing mechanism;
- DP8992 (Coaxial Transceiver Interface, CTI) receives and transmits data over a coaxial cable.
Interesting fact: the ubiquitous Campbell founded his own company, SEEQ Technology, for the production of Ethernet components, including 8992 controllers.
Later, Chipernet technology (as Ethernet was previously called) was supplemented with the ability to transmit data over unshielded twisted pair conductors UTP (Unshielded Twisted Pair). It is important to emphasize that Ethernet was intended to be a low-cost and effective alternative to other networking solutions. Therefore, it seems completely logical to expand capabilities using twisted pair cables.
One of the leaders in the production of inexpensive network controllers using Ethernet became the "Western Digital Corporation", better known as Western Digital. This happened at a time when hard drives had not yet become the “crown number” of WDC (later, due to a change in interests, the development of network technologies was sold to SMC). Since then, the famous trinity SMC, 3Com, Intel has long ruled the world of NE2000-compatible network adapters.
In the world of NE2000-compatible devices, three other companies placed emphasis: Realtek (60% of the market for all network controllers), VIA Technologies, Winbond Electronics. The latter is more familiar to consumers under the Compex brand. Practice
Three sources, three components...
At the rate of improvement of its characteristics, for example, an increase in the upper cutoff frequency transmission path and bandwidth, cable systems are practically not inferior to modern processors with their growing clock frequencies. This fact alone gives reason to assert that this area is one of the most dynamically developing on the market information technologies. As in any other area with high rates of development, this market has its own technical, organizational and marketing problems, and in the process of classifying the elements of a structured cabling system (SCS), into which a modern computer network “fits,” various, often irreconcilable, collide approaches and schools.
But no matter how many main groups and classes the “fathers of network engineering” would divide the components of a modern network, for the distribution of signals in it, in addition to the access devices responsible for physical interface, at least two more important parts are required that are involved in the formation of the physical transmission medium: cables (we will deliberately limit ourselves to considering the workstation subsystem and the horizontal “copper” subsystem) and connectors for connecting them. These components of a modern SCS have been described many times, but the need for a small “potpourri” on this topic is due to the fact that, for example, despite the general reduction in prices for fairly high-quality Cat.5e copper cables, users are often forced to impose a wide range of frankly “bazaar” products (suitable except for creating a home network structure). In a more serious case, this becomes one of the sources of constant headache for network maintenance personnel, who for the most part have to manage (alas!) without expensive professional network analyzers, allowing you to identify almost all network problems with one click of a button.
For basic UTP use, a single-core 4-pair cable with a conductor diameter of 0.51 mm (24 AWG) is specified. According to other canons, the use of a single-core cable with a conductor diameter of 0.64 mm (22 AWG) is also allowed. For a multi-core patch cord (UTP, the same 100 Ohms), the task of ensuring a long service life is urgent, despite frequent inevitable bends during operation. Let us note right away that despite a certain “loyalty” of the standards in relation to multi-core cables for crossover cords and connecting (user) cables (for them the standard allows 20-50% greater attenuation depending on which standard is followed - American or international ), in all other respects they must meet the minimum cable performance requirements of a horizontal system.
Performance labeling must be present to indicate the appropriate category. These labels should not replace safety class labels. As an example, let's look at the markings on the cable of our test system.
Cable marking
* NVP (Nominal Velocity of Propagation) nominal speed of propagation wave shortening coefficient in the cable. It shows how many times the speed of signal propagation over twisted pairs less speed light in a vacuum.
About color coding and correct termination
With this order of connecting pairs, as indicated in the table, the value and sign of the distribution of signal propagation delays guaranteed by the manufacturer are ensured.
Crimping options for RJ-45 connectors
Connector Termination Standards
Options "A" and "B"
The latter can be explained simply: in order to reduce crosstalk between pairs and eliminate possible resonance phenomena when unused pairs are not fully matched to the load (and in some network adapters we found only four contacts in the socket instead of eight), the conductors are twisted in pairs with different pitches (the number of twists per unit length). For the same reason, it is also advisable to take into account that the connection between the socket and the connector plug is made through eight closely spaced parallel contacts, which causes capacitive coupling between them. The degree of this influence also depends on the way the contacts are connected to the corresponding cable pairs (see figure). In the 568 A version, pair 2 is disconnected by pair 1, in the sequence 568 V pair 3 by pair 1.
The RJ45 standard (you can find the name of the 8P8C connector) came to the world of computer networks from telephony. It provides an asymmetrical detachable connection. The modular connectors of the RJ family are available in two versions, designed for cables with different core types. Looking ahead a little, we point out that flexible patch cords (flat modular two-, four-, six- or eight-core Cat.3 and four twisted pairs Cat.5) have a core consisting of several wires. Therefore, to make such cables it is necessary to use a connector with a contact that cuts into the core body. The installation cable has a core made of solid copper conductor, so connectors with split contacts are used to install these cables. Accordingly, if the connector is not intended for of this type cable, then it will not be possible to achieve high-quality contact.
There are several options for the relative position of the conductors relative to the connector contacts. To connect all four pairs of conductors (remember that Fast Ethernet uses two pairs for operation, you will need four when switching to a gigabit network) TIA-T568A, TIA-T568B are common (see table).
Connecting pairs to contacts that do not comply with the standards can lead to so-called split pairs, i.e. a situation where the connector is connected in such a way that the pair consists of wires from two different twisted pairs. This configuration sometimes allows network devices to exchange data, but often becomes the source of a difficult-to-diagnose problem - it is not only susceptible to excessive transient noise, but is also less resistant to external noise, including those that periodically appear due to the specific location of the cable. Result: errors during data transmission. These separated pairs allow cable testers to identify them.
In general, if we omit the previously made comments, it is permissible to use both of these options. However, here is a quote for those who are trying to perceive the table of options as a recommendation for making crossover cables: “...provided that both ends are terminated using the same wiring option.”
Patch cords: straight and crossover
Basic rules for laying cables
Some rules for installing UTP cable systems, the validity of which we have seen from our own experience.
- To avoid stretching, the tension force for 4-pair cables should not exceed 110 N (approximately 12 kg force). As a rule, a force exceeding 250 N leads to irreversible changes in the parameters of the UTP cable.
- The bending radii of installed cables should not be less than four (some manufacturers insist on eight) diameters for horizontal UTP cables. Allowable bending during installation is at least 3-4 diameters.
- Excessive stress on cables should be avoided, usually caused by twisting (the formation of “wings”) during pulling or installation, excessive tension on overhead sections of routes, tightly tightened narrow cable clamps (or “shot” staples).
- Horizontal system cables must be used in combination with switchgear and patch cords (or jumpers) of the same or higher performance ratings.
- And, perhaps, the most important thing to remember throughout all installation work is that the quality of the assembled cable system as a whole is determined by the line component with the worst performance characteristics.
Distribution panels and subscriber sockets
The patch panel is used for convenient and quick switching of various ports and equipment. With its help, you can instantly configure working ports for data, audio and video transmission. Horizontal cables run from workstation outlets to patch node patch panels, where they are exposed as user ports. The corresponding user ports can then be switched with LAN ports, video ports and telephone exchange ports. However, in a small network, the patch panel takes on a completely different meaning, serving mainly not so much as a means of streamlining network management and quick reconfiguration, but as a way to save yourself from additional problems during subsequent network modernization and expansion. It is clear that if, for example, the initially purchased hub is designed for 8 ports, and there are 12 computers in the office, then this is a “hassle.” At a minimum, you will have to buy another hub and cascade them, or at maximum, purchase a switch with 16 or even 24 ports. However, if initially a sufficiently “capacious” patch panel (for the same 16 or 24 ports) was used for switching, then it will be possible to avoid the much greater hassle of redrawing the cable management. Patch panels differ in the number of ports, standards, and switching method. In terms of the number of ports, the most common are 12-, 24- and 48-port. They typically have a mounting width of 19" (the form factor of most standard cabinets) and provide space for channel markings.
The next and most often visible element of the cable system from the customer’s point of view is the subscriber socket. The design of the module minimizes the installer's actions when connecting to the cable, allows maintaining the required bending radius of the cable, and does not require the use of any tools when placing the module in the box. The socket contacts can be additionally covered with a special curtain that prevents dust from getting inside.
Installation cabinets are designed to house switching and active equipment. Cabinets can be equipped with a cooling and ventilation system, glass and metal doors, a movable baseboard on four wheels with brakes, and locks on the door. There is usually enough space along the side walls of cabinets for wiring bundles and ventilation. However, for small networks, the installation cabinet is still more of an element of chic than a real necessity. Although if you have money and the desire to “do it beautifully”...
What tool might you need?
To work with UTP-type cables, a whole range of fairly convenient combined tools has been created that perform cable cutting, standardized ring trimming to remove the top insulation and stripping of individual cores (if this is required for this type of equipment, because modern installation methods based on mortise contact technology do not require stripping).
Without touching on the specialized tools and equipment recommended for terminating cable cores to patching and distribution panels (you can find them on the websites of their manufacturers), we decided to focus on a tool designed for “everyday” work, crimping a plug on an RJ-45 cable. Its numerous variants differ both in the range of functions performed and types of crimped connectors, and (quite significantly) in service life and price.
For minor repairs, you can try using an economical plastic tool. However, it is suitable only for a minimal amount of occasional installation work, and, as practice shows, to modernize a network with a hundred ports, its resource may be enough for no more than six months to a year.
The metal professional tool ensures that the punches move strictly perpendicular to the parting surface, which has a beneficial effect on the quality of work. As a rule, such tools have a multi-joint mechanism with a “ratchet” to reduce and normalize the force applied to the handles. Universal kits that allow crimping of various types of connectors may include replacement and additional dies and punches that expand the functionality.
An intermediate position in terms of quality and parameters is occupied by simple single-joint metal devices, which are quite widely represented on the domestic market. They have a simplified mechanical design and a limited (but still 3-10 times longer than a plastic) service life due to rapid wear of the punch. The versatility of such tools is ensured not by replaceable sets, but by the presence of several surfaces on their working parts (2 in 1 and 3 in 1).
Speaking of testing and monitoring...
We have no doubt that in a simple peer-to-peer network of five machines the task of daily in-depth statistical analysis and weekly preventive testing. However, an informal blitz survey conducted during the work on the article regarding monitoring, diagnostics and testing of participants divided network owners and administrators into several groups, allowing us to formulate two extreme points of view that are by no means technical or financial:
- Interest in conducting network analysis and audit is directly proportional to the number of served workstations and, regardless of the topology and tasks performed, asymptotically approaches zero (up to complete indifference) if the number of clients does not exceed 15-20. In this case, most often the main “tools” used throughout the life of the network are a primitive cable tester and mastery of utilities such as ping and tracert. True, some respondents in this group recognize the need to measure the quantitative indicators of the cable system at the time of commissioning.
- The other extreme is when a large and rich company buys expensive network management, diagnostics and testing tools, but its network administrators practically do not use them in their work or use some of the simplest capabilities contained in them due to the fact that they either “ there is no time”, or “everything works for us anyway”, and in general they do not understand “why they need this”, or on their hardware platform or in the existing configuration these tools periodically “freeze”, “not everything is shown” or “ "They're lying." I didn’t want to, but I have to add often this situation turns out to be due to the fact that the capabilities of the available tools... simply exceed the qualifications of those who use them.
At the same time, the concepts of diagnostics and network testing are often identified, which in fact is fundamentally wrong. But diagnostics is usually understood as measuring characteristics and monitoring network performance indicators during its operation, without stopping the work of users. Network diagnostics is, in particular, measuring the number of data transmission errors, the degree of load (utilization) of its resources or the response time of application software. That is, the work that, in our opinion, a network administrator should do daily.
Testing is the process of actively influencing a network in order to check its performance and determine potential transmission capabilities network traffic. As a rule, it is carried out to check the condition of the cable system (quality compliance with standard requirements), find out the maximum throughput, or evaluate the response time of application software when changing the settings of network equipment or the physical network configuration. It is usually recommended to make such measurements by disabling or replacing users working on the network with test agents, which, as a rule, in real life leads to a rather long blocking " normal operation office." In addition, the duration of the procedure depends on whether primary measurements and analysis of parameters are performed or comparison of some required parameters with the primary results of reference (passport, certification) tests. However, in any case, most often this leads to the fact that both the procedure itself and its performers become “unpopular” both among ordinary workers and among management.
Although this goes beyond the technical scope, I would also like to note that diagnosing or testing a network often directly depends on... the degree of experience of the network administrator. “Young and green”, as a rule, diagnose and test the network often and with pleasure, because at the same time they do not so much correct or prevent problems as engage in self-education. Subsequently, when all these “games” (like any others) become boring, only really serious problems in its operation can force the network administrator to begin the process of diagnosing. And finally, with the advent of truly serious experience, the network administrator again “returns” to diagnostics and testing, but not so much because of youthful enthusiasm and curiosity, but because of the understanding of the need to carry out this procedure from time to time as a preventative measure.
Glossary
TestingNetwork adapter(network card) expansion card installed in a workstation, server or other network device that allows data exchange across network environment. operating system controls the operation of the network adapter through the appropriate driver. The amount of adapter and system CPU resources involved may vary from implementation to implementation. Network cards usually have a chip (or a socket for installing it) of “removable” memory for remote boot (Remote Boot), which can be used to create diskless stations.
Collision(collision) distortion of transmitted data on an Ethernet network, which appears when several network devices attempt to transmit simultaneously. Collisions - common situations that occur during normal operation Ethernet networks or Fast Ethernet, but a sudden increase in their number may indicate problems with some network device, especially when this is not associated with an increase in network traffic as a whole. In general, the probability of packet collisions increases when new devices are added to the domain and segments are lengthened (increasing the physical size of the network).
Collision domain(competing domain) a set of devices competing with each other for the right to access the transmission medium. The signal propagation delay between any two stations that belong to a given area must not exceed a specified value (often called the collision domain diameter and expressed in time units). When a device is connected to a switch, the number of collision devices in the domain is always reduced to two.
Horizontal cable is intended for use in a horizontal subsystem in the area from switching equipment (for example, in a cross-connection on a floor) to information outlets (at workplaces).
Patch cable(crossover) and termination (user) cords usually also consist of four twisted pairs and are very similar in design to “regular” UTP cable used in a horizontal subsystem. The main differences between them are that to provide resistance to repeated bending and extend service life, the conductors are multi-core, and the insulation may have a slightly greater thickness than a horizontal cable (about 0.25 mm). The outer insulating shell is made of a material with increased flexibility. The same marking and identifying inscriptions and length marks must be applied to it.
Disposal of communication channel network utilization the percentage of time during which a communication channel transmits signals, or otherwise the proportion of the communication channel capacity occupied by frames, collisions and interference. The “Communication channel utilization” parameter characterizes the degree of network congestion and the efficiency of using its potential capabilities.
Switch(Switch) multiport device link layer, which establishes an address connection between the sender and the recipient during the packet forwarding based on the MAC address switching table built and stored in it. Simply put, the switch emulates a “direct” connection between the receiving and transmitting devices. However, we should not forget that some (most often primitive unmanaged) switches, when overloaded in the network, i.e. when the traffic passing through them exceeds their capabilities, can actually temporarily “turn” into hubs.
Autonegotiation Auto Negotiation A process initiated by network devices to automatically configure a connection to achieve the fastest overall speed in a given environment. The priorities are: 100Base-TX full duplex, 100Base-TX half duplex, 10Base-T full duplex and 10Base-T half duplex. Auto-negotiation is defined by the IEEE 802.3 standard for Ethernet and is completed in a few milliseconds.
Half duplex(Half Duplex) a mode in which communication is carried out in two directions, but at any given time data can be transmitted only in one of them. In a network (segment) based on hubs, all devices can operate only in half-duplex mode, in contrast to a network based on switches, which can transmit in both full-duplex and half-duplex modes.
Full duplex(Full Duplex) bidirectional data transmission. The ability of a device or communication line to transmit data simultaneously in both directions over a single channel, potentially doubling throughput.
Physical connection speed(Wire Speed) For Ethernet and Fast Ethernet, this value is usually given as the maximum number of packets that can be transmitted over a given connection. The physical connection speed in Ethernet networks is 14,880, and in Fast Ethernet networks 148,809 packets per second.
MAC address(MAC address Media Access Control address) a unique serial number assigned to each network device to identify it on the network and control access to the medium. For network devices, addresses are set at the time of manufacture (specified by IEEE), although they can usually be changed using appropriate software. It is precisely due to the fact that each network card has a unique MAC address that it can exclusively take packets intended for it from the network. If the MAC address is not unique, then there is no way to differentiate between the two stations. MAC addresses are 6 bytes long and are usually written in hexadecimal, with the first three bytes of the address identifying the manufacturer.
Test stand
Since such large-scale testing of network equipment is new for our laboratory (and, by the way, this topic is touched upon, frankly speaking, extremely rarely in other computer media), we took, so to speak, “the path of least resistance”, shifting maximum work is carried out by well-proven domestic suppliers of ready-made solutions and system integrators. Thus, the hypothetical “office computers” in our “reference LAN” are serial models of the Bravo PC from the K-Trade company, the server is really a server, specially selected through consultation with employees of the Kiev office of Intel and the system integrator Ulys Systems, and the cable hardware (patch cords with crimped connectors, patch cords, patch panel, etc.) was provided ready for deployment by ProNet.
For testing we used Bravo PCs with AMD processor Duron 1100 MHz, 256 MB PC133 SDRAM, AOpen AK73A motherboard (VIA Apollo KT133A), 40 GB HDD (Maxtor D540X), PowerColor GeForce2 MX400 video card (32 MB) and Windows 2000 Pro (SP3).
The server was a Dell PowerEdge 2500 system (Pentium III 1.26 GHz processor with the ability to install a second CPU; ServerWorks HE-SL chipset; 512 MB PC133 ECC SDRAM; Adaptec AIC-7899 Dual channel Ultra3 (Ultra160)/LVD SCSI controller; dual-channel SCSI RAID controller with a 128 MB cache buffer; three SCSI hard drives (10000 rpm), integrated into a RAID 5 array; integrated Intel PRO/100+ Server Ethernet controller; integrated video subsystem based on ATI-Rage XL 8 MB SDRAM; Server). This server configuration allowed us to get away from main problem the influence of the performance of the most “loaded” disk subsystem on the test results (after all, during many tests, all four PCs worked with the server simultaneously). The presence of a sufficiently high-performance processor and a relatively large amount of memory on the PC protected against the influence of unwanted side factors from “workstations”. The server and PC were controlled from a single operator console operating through a KVM-switch Raritan (provided by Ustar).
And this is what it all looked like assembled
To conduct tests of network adapters, a stand was assembled that allows simulating the operation of devices within the same collision domain. It is built using Molex Premise Networks horizontal LAN-level structured cabling equipment and includes four pieces of 2 x 15 m and 2 x 75 m Molex PN PowerCat.5E UTP cable connected to the 24-port patch plugs. Molex Cat.5E panels.
Stand layout
The cables were bundled and hung without boxes on hooks in the wall. As already mentioned, in electrically conductive systems it is necessary to take into account not only attenuation, but also interference. In our case, due to the fact that the cable fragments were folded in half during their installation, induced low-frequency interference from fluorescent lamps, lying in close proximity to power, signal cables, etc., as we expected, decreased (in-phase interference affecting the cable bundle).
In the process of creating the segment, it was decided to abandon standard subscriber sockets. To simulate their influence, we cross-crossed short (and, for reasons already explained above, extremely “harmful”) sections of the same cable, 8-10 cm long, onto patch panels.
Thus, instead of one pair of detachable contacts required for the completeness of the experiment, we were able to connect two more, including them in the open circuit from the hub to the machine with an additional patch cord. In the Test Laboratory, it is usually not customary to trust even well-known brands without the appropriate instrumental confirmation, so immediately after installation they not only checked the correct connection and distribution of cable cores, but also measured the quantitative parameters of each of the segments using a portable OMNIScanner II analyzer from Fluke Network.
Fluke OMNIScanner II in person
Indicators of the 75-meter segment
Indicators of 15-meter segments
Indicators of a short “bent” segment
Methodology
Since identical network cards were installed in turn on all four PCs, we were naturally interested in creating, if possible, different conditions for their functioning. Ultimately, we settled on the configuration that can be seen in the stand diagram two “long” segments of 75 and 90 meters, one “ideal connection” (the communication cable from the computer is directly connected to the hub) and one short “inconvenient” connection through a small piece of bent cable. Looking ahead, we note that our assumptions were largely confirmed - some network card models actually behaved differently depending on the segment length on which they had to operate. The server was “distanced” from the hub by 15 meters, which is quite consistent with the maximum of the actually encountered options (within reason).
Perhaps some will be surprised that we chose a hub rather than a switch as the device that unites network subscribers. The answer is quite simple: the fact is that to create a load for the actual tests, i.e., on network cards, a switch in a network of four clients and one server is simply unsuitable. In fact, we deliberately complicated the task by increasing the number of collisions in the network to the maximum level that could actually be achieved, in order to identify weaknesses in the operation of network controllers. If the switch were used, all tests would actually turn... into a study of the performance of the switch itself. A few words about the hub. Oddly enough, we chose a rather simple and cheap LG model, made based on Realtek chips. This happened for two reasons: firstly, companies like Intel, 3Com or Cisco have now practically abandoned the production of hubs, and secondly, routine tests using other models (3Com Office Connect and CompuShack 5DT Desktop) showed that Replacing this particular device in our case did not have any effect on the test results.
The tests included performance studies using the popular (as far as one can talk about the popularity of such software) package eTestingLabs NetBench 7.02 (modified script NIC_nb702, in which packet sizes were left at 512, 4K, 16K and 64K), CPU load measurements standard utility Windows 2000 Performance Monitor while copying a 512 MB file from one of the clients to the server, as well as measuring the speed of “counter” copying of two 1 GB files between two clients connected by a crossover cable (checking the correctness and effectiveness of the full-duplex mode).
Fast Ethernet Adapter Specifications
| Manufacturer | Model | LED indicators | Wake-On-LAN | IC Boot ROM | Network chip | Orient. price, $ | Warranty, years |
| 3Com | 3C905CX-TX-M | 10-100/Link/Activity | Connector/cable included | Preinstalled | 3Com 920-ST06 | 43 | 5 |
| Home Connect 3C450 | 10-100/Link/Activity | Not supported | Not supported | 3Com/Lucent 40-04834 | 22 | 1 | |
| Allied Telesyn | AT-2500TX | 10-100/Activity | Supported | Crib | Realtek RTL8139C | 13 | 1 |
| ASUS | PCI-L3C920 | Link/Activity | Not supported | Crib | 3Com 920-ST03 | 32 | 1 |
| CompuShack | Fastline II PCI UTP DEC-Chip | Link-FDX/Coll/SPD-100/Act | Connector/cable included | Crib | Intel (DEC) 21143-PD | 33,6 | 3 |
| Fastline PCI UTP Realtek-Chip | Link/Activity | Connector/cable included | Crib | Realtek RTL8139C | 11,2 | 3 | |
| D-Link | DFE-528TX | Link/Activity | Not supported | Not supported | D-Link DL10038C | 13,6 | Lifetime |
| DFE-550TX | Link/100/FDX | Connector/cable included | Crib | D-Link DL10050B | 22,3 | Lifetime | |
| Intel | InBusiness 10/100 | Link/Activity/100Tx | Not supported | Not supported | Intel GD82559 | 25 | 1 |
| Pro/100M Desktop Adapter | Link/Activity/100Tx | Not supported | Preinstalled | Intel 82551QM | 29 | Lifetime | |
| Pro/100S Desktop Adapter | Link/Activity/100Tx | Connector/cable included | Preinstalled | Intel 82550EY | 31 | Lifetime | |
| Lantech | FastLink/TX | 10/100/FDX/Activity | Connector/cable included | Crib | Intel (DEC) 21143-PD | 27 | 2 |
| FastNet/TX | Link/Activity/FDX | Not supported | Crib | Realtek RTL8139D | 6,5 | 2 | |
| LG | LNIC-10/100Aw | Link/Activity | Connector/cable included | Crib | Realtek RTL8139D | 6,2 | 1 |
| Planet | ENW-9504 | 10-100/Activity | Not supported | Not supported | Realtek RTL8139D | 9,5 | 3 |
| SMC | EtherPower II 10/100 | Link/FDX/Tx/Rx | Connector/cable included | Crib | SMC 83С172ABQF | 42 | 5 |
| Surecom | EP-320X-R | Link/Activity | Not supported | Crib | Realtek RTL8139C | 9 | 2 |
| EP-320X-S | Link/Activity | Not supported | Crib | Myson MTD803A | 8 | 2 |
Test results
First, let's explain why, despite testing network cards, you can only see the names of the chips in the diagrams. The fact is that despite our completely “honest” behavior, which was expressed in using not “generic” drivers from chip manufacturers, but the latest available versions from card manufacturers, there is no difference in performance between cards made on the basis of the same We didn’t find any microcircuits.
Typical "single chip" network card
NetBench test results are presented in a limited volume for one reason - in all other cases they were simply... exactly the same. Only the test with a packet size of 16K revealed some peculiarities in the functioning of our test network, and it was the difference in the results demonstrated by the network cards that interested us most of all. But this subtest more than paid for our expectations - the average throughput of each of the four clients sometimes differed several times! Having brought together all the “distinguished” chips and tried to find some kind of dependence, we noticed that the most significant results belong to network Intel controllers and 3Com, and this immediately brought us to one obvious thought...
Both one and the other company do not bother simple copying the long-known “exemplary diagram of a classic network chip”:
Additionally, they use so-called “adaptive technologies”, which allow them to regulate the amount of information transmitted on the network and the amount of latency in order to make the most of the capabilities of a particular environment and achieve the highest overall network throughput. It seems that in our case, cards located on “inconvenient” (or, for the sake of correctness, let’s make a reservation considered inconvenient according to the underlying analysis algorithm) segments, “voluntarily ceded” part of the strip to their counterparts located in better conditions. It should be noted that this still did not bring a gain in the total volume of transmitted data; if you add up all the throughput values for each of the clients, their sum will be approximately the same as in the case of more “straightforward” cards. In general, we will refrain for now from assessing this feature of some network chips at the “good/bad” level, because depending on the specific operating conditions of the network and the tasks being solved in it, it can easily change in each specific case to the diametrically opposite one.
Chips
3Com 920-ST06/03. A “smart” chip that clearly supports adaptation technologies to the conditions of a specific cable environment (enough has already been said above about the “ambiguity” of this approach). Demonstrates the lowest CPU load and decent support for full-duplex communication mode. A classic example of a good, but expensive solution.
3Com 3C905CX-TX-M
ASUS PCI-L3C920
3Com/Lucent 40-04834. There is also a very low processor load and decent support for full duplex mode, but somewhat more “moderate” intelligence - which, however, can sometimes be useful. But the cost of such a solution is two times lower than that of a newer one.
3Com Home Connect 3C450
D-Link DL10050B. But this is a classic example of a simple but high-quality chip no attempts to take into account the features of a specific line, but at the same time full duplex and the lowest CPU load among “second-level brands”. Conventionally, this chip, taking into account the price of the card based on it, can be called a simplified analogue of 3Com/Lucent 40-04834, equal to it in almost everything, but without adaptive properties and with a higher CPU load.
D-Link DFE-550TX
Intel (DEC) 21143-PD. A very ambiguous chip, however, given its age... Some “rudimentary” adaptive properties, but unexpectedly high load processor and a complete failure in the Full Duplex support test. It is worth mentioning one feature that we noticed during the tests: the card from CompuShack was at least able to complete the “counter copy” test, albeit with a worse result, but Lantech FastLink/TX in the middle of the test began to simply... regularly “lose” " net! In short, on the one hand, in systems based on hubs, where full-duplex mode support is not required, cards based on 21143-PD can be used, but on the other hand, this solution can hardly be called optimal.
CompuShack Fastline II PCI UTP DEC-Chip
Lantech FastLink/TX
Intel 82550EY. Another version of a “superintelligent” device, distinguished by its dislike for long segments. Full duplex support is excellent, CPU load is very low. In terms of all its properties, it is the closest competitor to the 3Com 920-ST06/03, but with a much more affordable price. What’s interesting is that there was once a case when one of the independent Western test laboratories conducted a comparative study of the performance of network Intel chips and 3Com, after which both companies, interpreting the same numbers in their own way... announced that, according to the results of these tests, their chip is better than that of a competitor!
Intel Pro/100S Desktop Adapter
(PCB for Pro/100 M and InBusiness 10/100 is similar)
Intel 82551QM (Intel card Pro/100M). Everything said above about the Intel 82550EY can be repeated in this case, but with one caveat: this chip “did not like” another segment of our test network. To be honest, for now we decided to simply present this as a fact, as they say, “as is,” since the behavior and preferences of chips that support adaptation technologies deserve a separate study.
Intel GD82559(InBusiness 10/100 card). This cheapest network solution from Intel has clearly been slightly “reduced in intelligence”, however, while maintaining all the other positive properties of the chips from this company. And even the load on the CPU has dropped, and support for full-duplex mode, on the contrary, is the best among all participants! Quite a good solution for an “ordinary” car, it seems to us.
Myson MTD803A. In terms of cheapness, products based on this chip clearly compete with those based on Realtek chips and, in general, quite successfully. The lowest processor load among cheap chips, the same quality of support for full duplex mode as the RTL8139C. However, in the latter, the Myson chip is still inferior to the new version of Realtek RTL8139D.
Surecom EP-320X-S
Realtek RTL8139C / D-Link DL10038C. We combined these chips together because, although formally they are different, they performed exactly the same. At the first glance at the results of tests for CPU load and Full Duplex support, we, without saying a word, said the same thing: “Realtek has not changed itself.” Remembering the classics of Soviet literature Ilf and Petrov, we can, paraphrasing their saying, say that “this chip has full duplex... somehow incomplete.” However, they do work... And they are inexpensive.
Allied Telesyn AT-2500TX
CompuShack Fastline PCI UTP Realtek-Chip
D-Link DFE-528TX
Surecom EP-320X-R
Realtek RTL8139D. In short, we can simply state that from the point of view of test results, this chip is the same RTL8139C, which was slightly “treated” to support full duplex mode, and Realtek engineers did not have enough to “reach” the dense cohort of more famous competitors. However, high CPU load, the eternal “sore” of this company’s chips, remained unchanged.
Lantech FastNet/TX
LG LNIC-10/100Aw
Planet ENW-9504
SMC 83С172ABQF(SMC EtherPower II 10/100 card). Low load CPU, high speed of full duplex mode, but with increasing segment length there is a slight decrease in speed. Overall, it’s a good-quality and fairly old network chip without any major complaints, and it does its job honestly. But I would like to see a slightly different price for a solution of this class...
SMC EtherPower II 10/100
Conclusion
Well, we hope that this material will appeal to “beginning administrators and those simply interested” we tried to organically combine both theoretical aspects and practical advice in it, and the results of testing the most common desktop-level network controllers on the market will not be superfluous for “a young man thinking about making a net out of something.” In general, it is worth noting that, of course, behind the scenes there was not only “no less,” but even many times more than can be found in this material. It’s not surprising that thick books and monographs are written about how to properly design and configure a network, but we only had a little over a dozen pages of the weekly at our disposal. Therefore, you should probably not consider this article as a universal self-sufficient guide or, God forbid, a textbook. The information that it contains, perhaps, may only be enough to understand a few simple truths: firstly “it’s not the gods who burn the pots”, and it’s quite possible to learn how to do some things on your own, secondly before doing this “something”, it is still advisable to obtain at least a basic set of knowledge about the subject, and thirdly, even having received this basic set, it’s clearly not worth stopping there. It is impossible to “know what a LAN is”, you can only study it. How many? Yes, even for the rest of your life!
Products provided by companies:
3Com Ingress, NIS
Allied Telesyn "ICS-Megatrade", ELKO Kiev
ASUS "Technopark"
Compu-Shack N-Tema, Service ASN
D-Link "Version"
Intel K-Trade
Lantech Compass, N-Tema
LG DataLux, K-Trade
Planet MTI, Engler-Ukraine
SMC "Ingress"
Surecom IT-Link
Cables used to build computer network infrastructure are available in a wide range of varieties. Among the most popular are coaxial, twisted pair, and optical fiber. What are the specifics of each of them? What are the features of installing the most common type - twisted pair?
Cable types: coaxial
Among the most historically early types cables used in network connections, - coaxial. Its thickness is approximately the same as that of a computer power supply, designed to work with a 220 V outlet.
The structure of the coaxial structure is as follows: in the very middle there is a metal conductor, it is wrapped in thick, most often plastic, insulation. On top of it is a braid of copper or aluminum. The outer layer is an insulating shell.
The connection of the network cable of the type in question can be done by:
BNC connector;
BNC terminator;
BNC-T connector;
BNC barrel connector.
Let's consider their specifics in more detail.
The BNC connector is supposed to be placed at the ends of the cable and is used to connect to T- or barrel connectors. A BNC terminator is used as an isolating barrier that prevents signal movement along the cable. Correct functioning of the network without this element is in some cases impossible. A coaxial cable requires the use of two terminators, one of which requires grounding. The BNC-T connector is used to connect the PC to the main line. There are three slots in its structure. The first one is connected to the computer connector; the other two are used to connect different ends of the line. Another type of connector for coaxial cable is the BNC barrel. It is used to connect different ends of a highway, or to increase the radius of a computer network.
Among the useful features of coaxial designs is that there are no problems with deciding how to connect two network cables of this type. It is enough to ensure reliable contact of the conductive cores, of course, subject to the technology of pairing the insulation and the screen mesh. However, coaxial cable is quite sensitive to electromagnetic interference. Therefore, in the practice of building computer networks, it is now used quite rarely. However, it is indispensable in terms of organizing the infrastructure for transmitting television signals - from dishes or cable providers.
twisted pair
Probably the most common network cables for computers today are called “twisted pair”. Why this name? The fact is that the structure of this type of cable contains paired conductors. They are made of copper. A standard cable of the type in question includes 8 cores (thus, 4 pairs in total), but there are also samples with four conductors. The so-called pinout of a network cable of this type (correlating each core with a particular function) involves the use of insulation of a certain color on each conductor.
The external insulation of the twisted pair is made of PVC, which provides sufficient protection of the conductive elements from electromagnetic interference. There are the types in question - FTP and STP. In the first, the foil performing the corresponding function is located on top of all the cores, in the second - on each of the conductors. There is an unshielded modification of twisted pair - UTP. As a rule, cables with foil are more expensive. But it makes sense to use them only if there is a need for high-quality data transmission over relatively long distance. For home networks, an unshielded twisted pair version is quite suitable.
There are several classes of the corresponding type of structure, each of them is designated as CAT with a number from 1 to 7. The higher the corresponding indicator, the better the materials that ensure signal transmission. Modern network cables for computers for data exchange via Ethernet in home networks require elements to comply with CAT5 class. In connections that use twisted pair, connectors are used that would be correctly classified as 8P8C, but there is also an unofficial name for them - RJ-45. It can be noted that cables that meet at least the CAT5 and CAT6 classes can transmit data at speeds close to the maximum for the type of structure under consideration - up to 1 Gbit/s.
Optical fiber
Perhaps the most modern and fastest network cables for computers are fiber optic cables. Their structure contains light-conducting glass elements, which are protected by durable plastic insulation. Among the key advantages of these network cables for a computer is high immunity to interference. Also, data can be transmitted over a distance of about 100 km via optical fiber. The connection of cables of the type in question to devices can be carried out using various types connectors. Among the most common are SC, FC, F-3000.
What does this high-tech network cable for a computer look like? Photo of the fiber optic structure below.
Intensity practical application Optical fiber is limited by the relatively high cost of the equipment required to transmit data through it. However, recently many Russian providers have been actively using this network cable for the Internet. According to IT experts, with the expectation that the corresponding investments will pay off in the future.
Evolution of cable infrastructure
Using the example of the three noted types of cables, we can trace some evolution in the aspect of building computer network infrastructure. Thus, initially, when transmitting data via the Ethernet standard, coaxial structures were used. At the same time, the maximum distance over which a signal could be sent from one device to another did not exceed 500 meters. The maximum over coaxial cable was about 10 Mbit/sec. The use of twisted pair cables has significantly increased the dynamics of file exchange on computer networks - up to 1 Gbit/sec. It also became possible to transmit data in duplex mode (one device could both receive signals and send them). With the advent of optical fiber, the IT industry was able to transfer files at speeds of 30-40 Gbit/sec or more. Thanks largely to this technology, computer networks successfully connect countries and continents.
Of course, when working with a PC, many other types of cables are used that are used in the installation of computer networks. Theoretically, for such purposes, you can use, for example, a USB cable, although this will not be very effective, in particular, due to the fact that within the USB standard, data can be transferred over a short distance - about 20 m.
How to connect twisted pair
Twisted pair, as we noted above, is today the most common type of cable in the design of computer networks. However, its practical use is characterized by some nuances. In particular, they reflect such an aspect as the pinout of the network cable, which we mentioned above. It is important to know how to correctly position the wires in the area where they come into contact with the RJ-45 connector. The procedure by which a twisted pair is connected to the corresponding element is called crimping, since during its implementation a special tool is used that involves force on the structure.
Nuances of crimping
During this procedure, the connectors are securely fixed to the ends of the twisted pair. The number of contacts in them corresponds to the number of cores - in both cases there are 8 such elements. There are several schemes within which twisted pair cables can be crimped.
Next we will look at the relevant specifics. But first, the person working with the cable needs to properly hold the connectors in their hands. They should be held so that the metal contacts are located on top.
The plastic latch should be directed towards the person doing the crimping. In this case, the 1st contact will be on the left, and the 8th contact will be on the right. Numbering is an extremely important aspect of working with twisted pair cable. So, what crimp schemes are used by network infrastructure specialists?
First, there is a network cable design called EIA/TIA-568A. It assumes the arrangement of the cores in relation to the metal contacts of the connector in the following order:
For 1 contact: white-green;
For the 2nd: green;
For the 3rd: white-orange;
For the 4th: blue;
For the 5th: white and blue;
For the 6th: orange;
For the 7th: white-brown;
For the 8th: brown.
There is another scheme - EIA/TIA-568B. It assumes the arrangement of the cores in the following order:
For 1 contact: white-orange;
For the 2nd: orange;
For the 3rd: white-green;
For the 4th: blue;
For the 5th: white and blue;
For the 6th: green;
For the 7th: white-brown;
For the 8th: brown.
You now know how to connect a network cable to a connector. But it is useful to study the specifics regarding various twisted pair connection schemes to certain devices.
Crimping and connection type
So, when connecting a PC to a router or switch, you should use the direct connection method. If there is a need to organize file exchange between two computers without using a router, then you can use the cross connection method. The difference between the marked schemes is small. At direct method connections, the cable must be crimped using the same pinout. When crossed, one end is according to circuit 568A, the other is according to 568B.
High-tech savings
Twisted pair is characterized by one interesting feature. With a direct connection scheme, the device can use not 4 pairs of conductors, but 2. That is, using one cable it is permissible to connect 2 computers to the network at the same time. This way you can save on cable costs or make a connection if you really need to do this, but you don’t have extra meters of twisted pair at hand. True, in this case the maximum data exchange speed will not be 1 Gbit/sec, but 10 times less. But for organizing homework, it is acceptable in most situations.
How to distribute the cores in this case? In relation to the contacts on the connection connectors:
1 contact: white-orange core;
2nd: orange;
3rd: white-green;
6th: green.
That is, 4, 5, 7 and 8 cores are not used in this scheme. In turn, on the connectors for connecting a second computer:
1 contact: white-brown core;
2nd: brown;
3rd: white-blue;
6th: blue.
It may be noted that when implementing a crossover connection scheme, you must always use all 8 conductors in a twisted pair. Also, if the user needs to implement data transfer between devices at a speed of 1 Gbit/sec, the pinout will need to be carried out according to a special scheme. Let's consider its features.
Gigabit speed cross-connect
The first cable connector should be crimped according to diagram 568B. The second assumes the following comparison of cores and contacts on the connector:
1 contact: white-green core;
2nd: green;
3rd: white-orange;
4th: white-brown;
5th: brown;
6th: orange;
7th: blue;
8th: white and blue.
The circuit is quite similar to the 568A, but the position of the blue and brown wire pairs has been changed.
Compliance with the marked rules for matching the colors of cores and contacts on the 8P8C connector - most important factor ensuring functionality network infrastructure. The person designing it must be careful when installing the relevant elements. It happens that the computer does not see the network cable - this is often due to incorrect crimping of the twisted pair cable.
How to crimp a cable correctly
Let's look at some technical nuances. The main device that is used in this case is a crimper. It is similar to pliers, but at the same time it is adapted to work specifically with computer cables of the appropriate type.
The design of the crimper requires the presence of special knives designed for cutting the structure. Also, sometimes crimpers are equipped with a small device for stripping twisted pair insulation. In the central part of the tool there are special sockets adapted to the thickness of the cable structure.
The optimal algorithm for the actions of a person crimping a twisted pair cable may be as follows.
- First of all, it is necessary to cut a section of the cable of suitable length - thus, precise measurements will be required.
- After this, the outer insulation should be removed - approximately 3 cm at the end of the cable. The main thing is not to accidentally damage the core insulation.
- Then you need to arrange the conductors in relation to the connection diagrams to the connector discussed above. Afterwards, trim the ends of the cores evenly, so that the length of each of them outside the outer layer of insulation is about 12 mm.
- Next, you need to put the connector on the cable so that the wires remain in the order that corresponds to the connection diagram, and each of them fits into the desired channel. You should move the wires until you feel resistance from the plastic wall of the connector.
- After the cores have been properly positioned inside the connector, the PVC sheath should be positioned inside the connector body. If you can't do this, you may need to pull out the wires and shorten them a little.
Once all the structural elements are positioned correctly, you can crimp the cable by inserting the connector into a special socket on the crimper and gently pressing the tool handle all the way.
If your mini-office does not require an extensive computer network, but you still need to connect 4-5 computers, you can do it yourself, especially since serious system integrators involved in organizing structured cabling systems (SCS) may not be interested in such a small project.
An office local network can either include a server and workstations, or be peer-to-peer (a network in which all computers are equal workstations). A server for a network of more than 5 computers is still preferable, since it allows centralized management network, and the software for the server machine provides some convenient features not provided in a peer-to-peer network.
To organize a local network, you will need a switch (hub, switch) with a sufficient number of network output connectors. Count the number of workstations and peripheral network devices (printers, faxes, etc.). It is better to buy a switch with a small margin in order to expand the network. If you plan to use a wireless connection, purchase a switch that has a built-in WiFi hotspot.
LAN cable
Currently, connecting computers into a network is carried out using: UTP cable of at least category 5.
For wiring to network sockets, a single-core cable is used; it is less flexible, but the thick cores of such a cable are well fixed in the socket connectors. To connect computers and peripheral devices to sockets, network hubs and connections between each other, a multi-core twisted pair (patch cord) is used. Although the signal attenuation in such a cable is somewhat greater, it is more flexible.
UTP cable has no protection against electromagnetic interference. It must be laid away from sources electromagnetic radiation. If this is not possible, you should use an FTP () cable (F/UTP, foil twisted pair). In case it is required reliable protection, they use a protected STP cable - not only does it have a common protective mesh, but each pair is protected by its own shield.
UTP cable can also be used to extend telephone lines in the office. But if for telephone lines a 3rd category cable is sufficient, then connecting computers requires no less than a 5th category, and better (for signal conductivity) - a sixth. For a twisted-pair local network, RG-45 standard sockets are used.
The UTP cable has limitations on the distance between the computer and the switch - no more than 100 meters.
When laying a single-core network cable, do not make sharp bends: the bend radius for a UTP category 5 cable should not exceed 8 diameters. It is better to purchase patch cords crimped (with connectors). When calculating the length of the patch cord, remember: its flexibility also has a limit and excessive twisting or stretching will lead to instability of the connection.
We will build modern LANs (LAN, WLAN) based on Gigabit Ethernet local area network network technology, providing a transfer speed of 1 Gb/s.
Network installation, basic principle.
Cabling.| These measures can reduce signal attenuation in the cable: | |
| as few bends as possible with a small radius of curvature; fewer connections; |
|
| The basic rules for laying cables are determined by the requirements of ISO/IEC 11801 and ANSI/TIA/EIA-568A standards.
Twisted pair wiring has a number of features. For example: a UTP twisted pair cable has a maximum permissible bending radius of eight outer cable diameters. Stronger bending leads to damage to the cable insulation and an increase in the intensity of external noise. |
|
When buying a cable (twisted pair), it is important to know!
Why is copper important? Examples.
* Categories UTP, FTP, STP twisted pair cables used in networks »
* Local network topologies »
* Types of passive fiber optic equipment »
Ethernet technologies in networks.
| Levels | Ethernet 10BASE-T | Fast Ethernet | Gigabit Ethernet |
| 1. End user (between the end user device and the device working group) | Provide connection: ▪ between end-user devices and user-level switches |
Provide high-performance (PC) server access at 100 Mbps | |
| 2. Workgroup level (connection of a workgroup device to the backbone) | At this level, as a rule, they are not used | Provide connection: ▪ between the end user and the working group; ▪ a block of servers and a backbone |
Provide high-speed links between: ▪ working group and highway; ▪ high-speed channels to a block of servers |
| 3. Trunk level | At this level, as a rule, they are not used | Connections to applications that require low to medium volume | Provide connections between high-speed highways and network devices |
| IN modern networks Although it is possible to provide Gigabit Ethernet connections from the backbone all the way to the end user, the cost of cables and switch ports may make such a solution impractical. Before making a decision in such a situation, it is necessary to correctly determine the needs of the network. For example, a network operating at traditional Ethernet speeds can easily become congested if it is running new generation software products such as multimedia, graphics applications, and database management systems. In general, Ethernet technologies can be used in territorial LAN networks in several ways below. |
At the user level, fairly high performance can be achieved using Fast Ethernet connections. Fast Ethernet and Gigabit Ethernet technologies can be used for clients or servers that require high bandwidth. |
Use the repeater as a signal amplifier!
| Amplified by the switch, data packets gradually fade and become distorted. Therefore, the signal can travel no more four switches. The maximum distance over which twisted pair cables can be laid using network switches is 900 meters. | |
 |
if the distance over which the network needs to be extended exceeds the capabilities of four switches, it is necessary to use a repeater. The repeater recalculates the data packets, which allows you to connect four more switches. using repeaters and network switches, you can lay twisted pair cable over an almost unlimited distance; in the diagram below, switch/hab are used, since they can also be used as a repeater. |
Scheme - we lay a LAN over 300 meters.
Scheme of a local area network for an office.
Checking (continuity) of the network cable.
If possible, lay the network cable in a metal or plastic pipe. This will protect it from accidental damage and rodents.
How to find the same cable, wire or core in a pile of wires?
 |
Analog tone generator for sending tones and wire tracking on inactive networks, and in particular for identifying pairs using SmartTone® technology. Clamps with angled piercing pins make wires easier to access, and the RJ-11 connector is ideal for use in telephone jacks. The powerful speaker on the network sensor allows you to hear sound through drywall, wood and other obstacles, which makes locating wires faster and easier. You can send a powerful tone signal over a distance of up to 16 kilometers - virtually any cable! |
| A tone generator is convenient when a huge network is being laid, where there are a lot of different cables and wires, which, in addition, also need to be labeled. This is where this “magic” device comes to our aid! | |
Twisted pair crimping.
 |
 |
 |
 |
| rj-45 cat.5 | crimper | stripper | twisted couple |
  |
 |
||
| rj-45 cat.6 (23AWG) | rj-45 cat.6, 6a, 7 (22/23AWG) | ||
Compressed straight patch cord (Internet over two twisted pairs)
 |
 |
 |
 |
 |
 |
 |
 |
Crossover in computer networks (crossover)- a crimped patch cord according to the EIA/TIA-568A - EIA/TIA-568B scheme for direct connection of network cards (network card, network adapter or Ethernet adapter) of two computers, i.e. is a patch cable or patch cord that connects two computers on a local network directly to each other without the use of network concentrators (hubs).
Wireless technologies are certainly beginning to take the lead in organizing home and office networks, but they will never replace wired ones, and if this happens, it will not be in the near future. The majority of providers install their cables directly to the client, and only then does the WiFi work routers. There are several types of LAN cables, they differ throughput channel, method of connecting to a computer, installation method and others. Let's consider in order how the standards changed, what they were, and what is used today.
What kind of cable is there for a local network?
The choice of conductor is initially always determined by the LAN topology, and the most common are coaxial wires and twisted pair. Fiber optic technologies are now widely used, but this is still a developing standard, used mainly for laying highways on long distances. It is not widely used for connecting the end user. So, LAN cable Ethernet comes in two types:
- Coaxial - which is a single-core wire with a screen, separated from each other by insulating material or an air gap. Reminds me very much TV wire resistance 70 Ohm.
- “Twisted pair” - consists of eight wires intertwined in pairs. Each core is marked with a separate color to simplify installation. The colors are fixed and described by specifications and all product manufacturers adhere to these rules.
- Fiber optic or fiber optic - has a very complex design and is quite expensive to install. The signal in it is transmitted in the form of light pulses through special light guides.
In the 90s of the last century, exclusively coaxial wire was used to build local networks, and on its basis such well-known topologies as “bus” and “ring” were developed. A little later, the “star” topology based on twisted pair cables appeared, which is still the most popular and popular architecture for local and global networks. Now it's time to stop and describe each one used LAN cable separately.
Coaxial cable and connectors used
This type of wire is the oldest of the conductors. This cord has one supporting copper or aluminum core, which is covered with a thick layer of insulating material. Next comes the screen, made in the form of a tape encircling the central core of aluminum or copper. The top outer layer is a sheath that protects the cores from damage, made of polyethylene or polyvinyl chloride. There are several types of such cable used for LAN:
- 10Base 5 is a thick type of conductor, with a cross-section of 12 mm and a total resistance of 50 Ohms for the 8th category and 75 Ohms for the 11th. The data transfer speed does not exceed 10 Mbit/s over distances between end nodes up to 500 meters.
- 10Base 2 is thin, about 6 mm in diameter, most common for organizing home or small office networks. Its resistance is 50 ohms, but the maximum length is 185 meters at a speed of 10 Mbps.
Thanks to good insulation, the signal in the conductor is practically not extinguished, i.e. packets are not lost, and additional algorithms for checking transmitted or received information are not needed. The only drawback is the rather high cost of production and low speed, so it was subsequently supplanted by “twisted pair”.
“Twisted pair” - types and methods of crimping
LAN cable“twisted pair” got its name due to the fact that it consists of eight strands intertwined in pairs. Each core is insulated in a strictly specified color. Polyvinyl chloride or polyethylene is used as an external insulating material that protects the signal from electromagnetic interference. There are several types of such cable:
- UTP (Unshelded Twisted Pair) is an unshielded modification, most often used for laying home or office networks when there is no strong interference with the transmitted signal.
- FTP (Foiled Twisted Pair) - a cable with an additional aluminum foil shield under the outer insulation.
- STP (Shelded Twisted Pair) - in addition to shared screen has an additional one, for each pair separately.
Twisted pair has 7 categories, and the higher the category number, the more protected the wire is from electromagnetic radiation. For Ethernet networks, Category 5 (CAT5) cable is used, which has a bandwidth of 100 MHz. When installing new facilities, it is recommended to use a more advanced modification of CAT5e for higher frequency signals with a bandwidth of 125 MHz.
Twisted pair is used to create connections at speeds from 100 Mbit/s to 40 Gbit/s, depending on the category and quality of the cable itself and its length between end devices. Typically the segment length should not exceed 100 m.
The wire must have a connector for connecting to network devices. For twisted pair, an RJ-45 connector is used (8P8C - 8 position and 8 contacts). Inside the connector there are special grooves with contacts for each core. There are several options for crimping network cables: forward and reverse (crossover). Straight patch cord is used to connect a computer to a router or switch or to connect active network equipment to each other. Cross is used quite rarely and serves to connect two computers to each other. Currently, almost all manufacturers install “smart” network cards, which do not care about the location of the wires in the connector, but it is advisable to adhere to the standards and arrange the wires as written in the specification. This will help avoid collisions in the operation of the entire network. To make a straight LAN cable The cores are arranged in the following order at both ends:
- white-orange;
- orange;
- white-green;
- blue;
- white-blue;
- green;
- white-brown;
- brown.
To make a patch cord for ease of work, special pliers are used - a crimper (or “crimp” in common parlance). The crimper allows you not only to clamp the wire evenly in the connector, but also to properly cut and strip the insulation. In exceptional cases, you can use a flat-head screwdriver or a knife, but then the quality will not be satisfactory. There are connectors that are clamped without a crimper, but they are designed for multi-core soft wires and may not be suitable for standard twisted pair cables.
Fiber optic cable
Optical fiber is the most advanced technology for transmitting signals to long distances at great speed. The difference in signal transmission is that light, rather than electricity, is used as an impulse. Light is transmitted through glass fiber strands and reflected from the inner walls of the conductor. You can simultaneously transmit several signals: they will not intersect or cancel each other. The speed of information transfer over such a cable is limited only by the capabilities of the network cards or adapters themselves. The cable is not subject to interference and is made of non-flammable materials.
The cost of such a cable is relatively low compared to other conductors, but its installation can only be carried out by qualified employees using high-precision and expensive equipment, so it is almost impossible to use it at home. But such a conductor has found wide application for laying highways, because the distances between signal amplifiers can reach hundreds of kilometers. Some providers already provide a service for connecting optical fiber to the home, but end devices are still connected via twisted pair, which is why it is the fundamental standard for organizing a network.
