Network technologies of information transmission. Technologies for transferring information in SAN Technologies for transferring data and information

There are the following information transfer technologies in computer networks: Fast Ethernet, IEEE 1394 / USB, Fiber Channel, FDDI, X.25, Frame Relay, ATM, ISDN, ADSL, SONET. The first four data transmission technologies: Fast Ethernet, IEEE 1394 / USB, Fiber Channel and FDDI are referred to as technologies local area networks... The rest were created for global communication channels. Let's consider some of the common data transmission technologies - Fast Ethernet, Fiber Channel, FDDI, ISDN.

Fast Ethernet or " 100Base-T"Is a high-speed data transmission technology in local networks. Data transfer rules using this technology are defined by the IEEE 802.3u standard. This standard describes the rules of operation of the second layer protocols of the OSI model (data link layer) and provides the ability to transfer data at a rate of 100 Mbps.

100Base-T technology uses CSMA / CD as a media access control protocol. 100Base-T builds on the scalability provided by the CSMA / CD method. Scaling implies the ability to continuously increase or decrease the size of the network without significantly reducing its performance, reliability and manageability. 100Base-T technology uses UTP5 (Category 5 Unshielded Twisted Pair) cable.

100Base-T technology has the following features.

1. Due to the use of the same media access control protocol - CSMA / CD networks, using 10Base-T Ethernet technology, are easily transferred to a higher-speed 100Base-T technology. Therefore, many manufacturers produce network cards supporting both data transmission technologies: 10Base-T Ethernet and 100Base-T. These network cards have built-in capabilities automatic detection the speed of data transmission in the network and automatic adjustment to the appropriate operating mode. Since 10Base-T Ethernet and 100Base-T can easily coexist on the same network, administrators have a very high degree of flexibility in migrating stations from 10Base-TEthernet to 100Base-T.
2. UTP5 cable and 100Base-T network cards are currently produced by a huge number of manufacturers.

The disadvantages of using 100Base-T technology are significantly greater restrictions on the length of cable segments than in 10Base-T Ethernet technology. Compared to 10Base-T Ethernet, which allows networks with a maximum diameter of 500 m, 100Base-T limits this diameter to 205 m. Existing networks that exceed this limit will require additional routers.

The promise of 10Base-T technology is that the new Gigabit Ethernet technology (also known as 1000Base-T or IEEE 802.3z) is being developed to accommodate existing UTP5 cabling systems. With this technology, the speed of data transmission in the network is increased to 1000 Mbps, which is ten times faster than data transmission using 100Base-T technology.

One of the relatively new technologies for data transmission is the Fiber Channel.

Technology Fiber Channel is based on the use of optical fiber as a data transmission medium. The most common application of this technology today is in high-speed storage area networks (SANs). Such devices are used to build high-performance cluster systems. Fiber Channel technology was originally created as an interface that allows high-speed data exchange between hard drives and a computer's processor. Later, the standard was supplemented and now defines the mechanisms of interaction not only between data storage systems, but also ways of interaction of several nodes of a cluster system with each other and data storage facilities.

Fiber Channel technology has several advantages over other media, the most important of which is speed. Fiber Channel technology provides data transfer rates of 100 Mbps. The second important advantage is the ability to transmit the signal over very long distances. Data exchange using a light signal instead of an electric one makes it possible to transmit information over distances of up to 10-20 km without using repeaters (when using a single-wave cable). The third advantage of Fiber Channel technology is complete immunity to electromagnetic interference. This quality makes it possible to actively use the optical transmission medium even in industrial premises with a large amount of electromagnetic interference. The fourth advantage is the complete absence of signal radiation to the environment, which makes it possible to use Fiber Channel in networks with increased security requirements for processed and stored data.

The main disadvantage of Fiber Channel technology is its cost: an optical cable with all the connectors and installation methods that accompany its use is significantly more expensive than copper cables.

For the organization of high-speed local area networks, FDDI (Fiber Distributed Data Interface) is used.

Technology FDDI is intended not for direct connection of computers, but for building high-speed backbone communication channels (backbone), uniting several segments of the local network. The simplest example of such a backbone is two servers connected by a high-speed communication channel based on two network cards and a cable. Just like 100Base-T technology, FDDI provides data transfer rates of 100 Mbps.

The FDDI network uses a dual physical ring topology. The transmitted signals move along the rings in opposite directions. One of the rings is called primary and the other is called secondary. With the correct functioning of the network, the primary ring is used for data transmission, and the secondary one acts as a spare.

In an FDDI network, each network device (network node) acts as a repeater. FDDI supports four kinds of nodes: station with double connection(DAS - dual-attached stations), single-attached stations (SAS), dual-attached concentrator (DAC), and single-attached concentrator (SAC). DAS and DAC always connect to both rings, while SAS and SAC only connect to the primary ring.

If a cable break or other breakdown occurs at any point in the network, making it impossible to transfer data between neighboring network nodes, then the DAS and DAC devices restore the network operability by redirecting the signal bypassing the inoperative segment using a secondary ring.

FDDI uses an access token as a media access control protocol and an optical cable as a transmission medium.

FDDI technology has the following advantages.

The dual physical ring topology ensures reliable data transmission by keeping the network up and running in the event of a cable break. The FDDI standard contains network management functions. In addition to the listed advantages, there is a specification (CDDI - Copper Distributed Data Interface) for building a network using FDDI technology using a copper twisted pair. This specification reduces the cost of network deployment by using less expensive copper instead of fiber.

The main disadvantage of FDDI is the cost of building the network. Network cards and optical cable for FDDI are significantly more expensive than other technologies that provide the same data transfer rate. The specifics of installing an optical cable requires additional training of specialists who work with the cable. Although CDDI NICs are cheaper than FDDI NICs, they are nevertheless more expensive than 100Base-T NICs.

Digital data exchange technology using telephone lines Integrated Services Digital Network (ISDN) provides the ability to exchange data in the form of digital signal transmission over digital telephone lines. This data can be a combination of video, audio, and other data. ISDN has several technological solutions that provide the customer with the required communication channel performance. For individuals and small offices, lines with a Basic Rate Interface (BRI) are mainly provided. For large companies, Primary Rate Interface - PRI lines are provided. BRI uses two 64 kbps bearer (B) links to receive and transmit data and one control channel (delta - D) to establish and maintain a connection. PRI is a collection of several digital lines used in parallel to receive and transmit data. Such sets of lines received the symbols T1 and E1. In the United States, the standard is the use of Tl lines. T1 consists of 23 B-channels and one D-channel with a total bandwidth of 1.544 Mbps.

E1 lines are used in Europe. E1 consists of 30 B-channels and one D-channel with a total bandwidth of 2.048 Mbps.

ISDN requires special equipment, including digital telephone lines, and network termination units (NT-1). The NT-1 converts the input signal to digital, distributes it evenly across the transmission channels and performs a diagnostic analysis of the status of the entire data line. NT-1 is also a point of connection to the digital network of various equipment: telephones, computers, etc. Also NT-1 can act as a converter for connecting equipment that does not independently support ISDN.

The advantages of ISDN are as follows.

1. The speed of data exchange has been increased with additional capabilities for integrating data, voice and video into a single stream.
2. Using ISDN, you have the ability to transmit data and voice traffic simultaneously over one telephone line.

The disadvantage of ISDN is its slow expansion due to the need to transform the existing telephone network infrastructure, which inevitably entails significant costs.

1. The subject of the discipline, the task and purpose of teaching the discipline
The discipline "Information transmission technologies" is one of the normative disciplines, which is included in the cycle of natural science (fundamental) training of specialists in the direction of "Computer Science".

The discipline provides for the consideration of the basic technologies for transmitting information in computer networks at the physical, channel and network levels.

The lecture material discusses telecommunication technologies, the main elements of information theory, characteristics and classification of information networks, a reference model (OSI), communication lines and data transmission channels, data transmission technologies at the physical layer, data transmission technologies at the data link layer in local and global networks, information transfer technologies at the network level in IP networks.

The purpose of the discipline:

familiarization with the basic elements of information theory and telecommunication technologies;
the formation of theoretical knowledge in the field of information transfer technologies in computer networks;
to teach to make a reasonable choice of the required technologies and means of information transmission in the development of computer networks and Web-applications;
to gain practical skills in working with the means of information transmission in computer networks at the physical, channel and network levels.

The task of studying the course "Information transfer technologies" is the theoretical and practical training of future specialists on such issues as:

information transfer technologies in computer networks;
information transfer protocols in LAN, dedicated (serial) communication lines and global networks with circuit and packet switching;
means of information transmission in information networks;
architecture of information networks.

2. What the student should know, be able to and what to be familiarized with as a result of studying the discipline Due to the study of the discipline, the student must
KNOW:

basic elements of information theory;
basic modern technologies information transfer at the physical, channel and network levels;
types and characteristics of communication lines and information transmission channels;
methods of converting signals and methods of multiplexing communication channels;
modern methods information transmission in composite networks.

substantiate the choice of information transfer technologies for solving practical problems in the process of designing computer networks;
to design the cable structure of a computer network;
to select the equipment of the cable system for building the LAN infrastructure.

BE AWARE:

with the main trends in the development of information transfer technologies;
with the prospects for the development of telecommunication technologies;
with modern means of information exchange and processing in local and territorial networks;

The curriculum of the course of 150 academic hours consists of two informative (educational) modules with a volume of 5 credits (the amount of ECTS credit is 30 academic hours) and consists of classroom studies and independent work of students.

Sources of information used:

Computer networks. Principles, technologies, protocols: Textbook for universities. 4th ed. / V.G. Olifer, N.A. Olifer - St. Petersburg. Peter, 2010 .-- 944 p.
Broido V.L. Computing systems, networks and telecommunications: Textbook for universities. 2nd ed. - SPb .: Peter, 2006 - 703 p.
Tkachenko V.A. that in. Comp "Uterine Merezhi and Telecommunications: Navch. Booker / V. A. Tkachenko, O. V. Kasilov, V. A. Ryabik. - Kharkiv: NTU" KhPI ", 2011. - 224 p.
A. L. Dmitriev. Optical information transmission systems / Textbook. - SPb: SPbGUITMO, 2007 .-- 96 p.

Local and global computer networks and technologies for their use in teaching schoolchildren

The modern system of general secondary education, all educational areas included in it, one way or another, are aimed at developing students' skills to work with information. It is no coincidence that in most state programs that determine the priority directions of the development of education in the Russian Federation, special attention is paid to the formation of general educational and general cultural skills of students working with information and means of processing it, which becomes the main core of the professional activity of graduates of educational institutions in an information society, a necessary component of information culture. ... In turn, the desire to form an information culture among future graduates leads to the orientation of general education towards the acquisition by students of knowledge about telecommunications and the media, the use of telecommunications for the acquisition of various knowledge and creative expression, the assessment of the reliability of information, the development of critical thinking, the correlation of information and knowledge, ability to properly organize the information process, evaluate and provide information security.
Telecommunication systems are of paramount importance not only in the system of general secondary education, but play a fundamental role in almost all spheres of society. At the level of development of the telecommunications information space, the most significant imprint is imposed by the level of development of primary communication networks and the level of development of network information technologies, which can rightfully be considered as technologies. transmission of information.
Under communication network they understand the totality of wired, radio, optical and other communication channels, specialized channel-forming equipment, as well as communication centers and nodes that ensure the functioning of a given network. In almost all modern communication networks used in the creation of information telecommunication systems, several network sections that are different in their characteristics are simultaneously present and work together. These circumstances largely determine the strategy and tactics of creating and using network information technologies.
Network information Technology developed simultaneously with the development of communication channels. At the beginning of the last century, the basis of telegraph and telephone communication networks was made up of analog wire and radio telecommunication channels, which then, with the development of microelectronics, began to increasingly be replaced by digital fiber-optic communication lines with significantly higher characteristics in terms of the quality and speed of information transfer. The concept of telecommunication technology has emerged, which unites the methods of rational organization of the work of telecommunication systems.
Telecommunication systems used today in the system of general secondary education, as a rule, are based on various connections of computers with each other. Connected computers can be viewed from different perspectives. On the one hand, the interconnection of computers is computer network... On the other hand, it is a means of transmitting information in space, a means of organizing communication between people. It is thanks to this property that computer networks are increasingly called telecommunication networks, thereby emphasizing their purpose, and not the features of their device.
Distinguish

· Local and global telecommunication networks. As a rule, a local network is called a network that connects computers located in one building, one organization, within a region, city, country. In other words, most often a local network is a limited space. Local networks are common in the field of education. Most schools and other educational institutions have computers connected to a local area network. At the same time, modern technologies make it possible to link individual computers located not only in different rooms or buildings, but located on different continents. It is no coincidence that you can find educational institutions that have branches in different countries, whose computers are connected to local networks. Moreover, local networks can also unite computers of different educational institutions, which allows us to speak about the existence of local networks in the education sector.
Unlike local networks, global networks have no spatial restrictions. Any computer can be connected to the global network. Anyone can access information posted on this network. The most famous example of a global telecommunications network is the Internet (INTERNET), which is being accessed by an increasing number of secondary schools. The Internet is not the only global telecommunications network. There are others such as the FIDO network or the SPRINT network.
Thus, most schools and other educational institutions of the general secondary education system have both local networks and the ability to use global networks.
With all the variety of information and telecommunication technologies, as well as ways of organizing data when sending them through communication channels, the world information computer network Internet occupies a central place. Moreover, today it is practically the only global telecommunications network that is universally used in the general secondary education system. This is largely due to the high speed and reliability of data transmission over the Internet in various formats (text, graphics, sound, video, etc.). The Internet provides an opportunity for collective access to educational materials, which can be presented both in the form of simple textbooks (electronic texts), and in the form of complex interactive systems, computer models, virtual learning environments, etc.
The number of users and sources of information on the Internet is constantly increasing. In addition, the quality of the telecommunication services provided is constantly improving. Thanks to this, high-quality Internet access is received not only by enterprises and organizations working in the economic and other spheres, but also by institutions of general secondary education.
The modern Internet is characterized by the presence of a serious problem of organizing a global search for information. The so-called search engines have been developed, which by the right word or a combination of words find links to those pages on the network in which this word or combination is presented. At the same time, despite the existence of existing search engines, the user has to spend a lot of time both on the process of searching for information and on processing and systematizing the data obtained.
In education this problem felt especially acutely: if educational information resources are presented on the network, then, as a rule, they are presented non-systematically. The lack of a systematic approach to the placement of such resources, as well as the lack of uniformity in solving psychological, pedagogical, technological, aesthetic, ergonomic and a number of other problems in the development and operation of educational resources of the Internet leads to the practical non-use of the advantages of telecommunications in order to improve the quality of the educational process.
The most widespread communication technology and the corresponding service in computer networks has become the technology of a computer method for sending and processing information messages, which provides operational communication between people. Email (E-mail) - a system for storing and sending messages between people who have access to a computer network. Any information ( text documents, images, digital data, sound recordings, etc.). Such a service department implements:

editing documents before transmission,
storage of documents and messages,
forwarding correspondence,
checking and correcting transmission errors,
issuance of confirmation of receipt of correspondence by the addressee,
receiving and storing information,
viewing received correspondence.

E-mail can be used to communicate with participants in the educational process and send educational materials. An important property of e-mail, which is attractive for general secondary education, is the possibility of implementing asynchronous information exchange. To use e-mail, it is enough to master several commands of the mail client for sending, receiving and processing information. Note that when communicating via e-mail, more psychological and pedagogical problems arise than technical ones. The fact is that in direct human communication, information is transmitted not only with the help of speech, other forms of communication are included here: facial expressions, gestures, etc. Of course, you can use "emoticons" to convey emotions during correspondence, but this does not solve the problem of impersonal communication. However, the transition to written language fosters positive traits such as accuracy, brevity of expression, and neatness.

E-mail can be used by educators for consultation, submission of tests and professional communication with colleagues. It is also advisable to use it for conducting an electronic lesson in an asynchronous mode, when the text of the lesson in electronic form, excerpts from the recommended literature and other training materials are sent to the students, and then consultations are held by e-mail.
A distinctive feature and convenience of e-mail is the ability to send the same message to a large number of recipients at once.
A similar distribution principle is used by another Internet service called mailing lists ... This service works in subscription mode. By subscribing to the mailing list, the subscriber at regular intervals receives mailbox a selection of emails on the selected topic. Mailing lists perform the functions of periodicals on the Internet.
In the general education system, using mailing lists, it is possible to organize the so-called "virtual classrooms" ... In the created study group of schoolchildren, the rules and methods of subscription are explained, and she begins to work. Every message addressed to the group by any of its members is automatically sent to all members of the group. One of the members of such a group may be a teacher.
The main didactic possibilities of using mailing lists are the automatic distribution of educational materials and the organization of virtual classrooms.
Teleconferencing is another popular service provided by modern telecommunication networks and implementing the exchange of information between people united by common interests.
Teleconference is an online forum organized for discussion and exchange of news on a specific topic.
Teleconferencing allows you to post messages of interest to dedicated computers on the network. Messages can be read by connecting to a computer and choosing a topic for discussion. Further, if you wish, you can reply to the author of the article or send your own message. Thus, a network discussion is organized, which is of a news nature, since messages are stored for a short period of time.
The presence of audio and video equipment (microphone, digital video camera, etc.) connected to a computer makes it possible to organize computer audio and video conferences, which are increasingly widespread in the system of general secondary education.
Unlike email-based distribution lists, some newsgroups and newsgroups are real-time. The difference is that in the case of a mailing list, information is exchanged off-line by automatic mailing emails... The news server publishes all messages on the general board immediately, and stores them for some time. Thus, teleconferences allow organizing discussion both on-line and in delayed mode. When organizing training sessions, it is advisable to use teacher-moderated newsgroups.
With development technical means computer networks are increasing the speed of data transmission. This allows users connected to the network not only to exchange text messages, but also to transmit audio and video over a considerable distance. One of the representatives of programs that implement communication over the network is the NetMeeting program, which is included in the set Internet Explorer... MS NetMeeting is a means of informatization that implements the possibility of direct communication over the Internet.
It should be noted that for the implementation of audio communication, appropriate technical equipment is required: a sound card, a microphone and acoustic systems. To transfer video, you need a video card and a camera, or only a camera that supports the Video for Windows standard.
The main directions of using MS NetMeeting in the educational process are:

organizing virtual training sessions and real-time consultations, including voice communication and video transmission of participants;
information exchange in text and graphic mode;
organization working together with educational information on-line;
sending educational and methodological information in the form of files in real time.

One of the most important telecommunication technologies is distributed data processing... In this case personal computers are used at the places of origin and application of information. If they are connected by communication channels, then this makes it possible to distribute their resources to separate functional areas of activity and change the technology of data processing in the direction of decentralization.
In the most complex systems of distributed data processing, connection to various information services and systems is carried out general purpose(news services, national and global information retrieval systems, databases and knowledge banks, etc.).
An extremely important service for general secondary education, implemented in computer networks, is automated information retrieval... Using specialized tools - information retrieval systems, you can quickly find the information of interest in world information sources.
The main didactic goals of using such resources obtained through telecommunication channels in teaching schoolchildren are the communication of information, the formation and consolidation of knowledge, the formation and improvement of skills and abilities, the control of assimilation and generalization.
The use of educational information resources available today, most of which are published on the Internet, allows:

organize various forms of activity of schoolchildren for the independent extraction and presentation of knowledge;
"apply the whole range of possibilities of modern information and telecommunication technologies in the process of performing various types of educational activities, including such as registration, collection, storage, processing of information, interactive dialogue, modeling objects, phenomena, processes, the functioning of laboratories (virtual, with a remote access to real equipment), etc .;
to use in the educational process the possibilities of multimedia technologies, hypertext and hypermedia systems;
diagnose the intellectual capabilities of schoolchildren, as well as the level of their knowledge, abilities, skills, the level of preparation for a specific lesson;
manage learning, automate the processes of monitoring the results of educational activities, training, testing, generate tasks depending on the intellectual level of a particular student, the level of his knowledge, abilities, skills, and the characteristics of his motivation;
create conditions for the implementation of independent educational activities of schoolchildren, for self-study, self-development, self-improvement, self-education, self-realization;
work in modern telecommunication environments, provide information flow management.

Thus, computer telecommunications is not only a powerful teaching tool that allows you to teach how to work with information, but, on the other hand, computer telecommunications is a special environment for people to communicate with each other, an environment for interactive interaction between representatives of various national, age, professional and other groups. users regardless of their location.
Unfortunately, many existing methods of effective use of telecommunication technologies in the process of teaching schoolchildren are still not fully used by teachers. A modern teacher, in addition to the ability to work with the latest computer technologies, should have an idea of the possible ways of using them in the educational process. The experience of theoretical and practical mastering by teachers of various methods of using telecommunication technologies in the learning process could become the basis for increasing the efficiency and quality of education, for the formation and further improvement of their professional skills.

Modern information transmission systems - ϶ᴛᴏ computer networks. The totality of all subscribers of a computer network is called a subscriber network. Communication and data transmission facilities form a data transmission network (Fig. 2.1).

Rice. 2.1 - Block diagram of a computer network.

The data transmission network consists of many geographically dispersed switching nodes connected to each other and to network subscribers using various communication channels.

A switching node is a complex of hardware and software that provides switching of channels, messages or packets. In this case, the term “switching” means the information distribution procedure, in which the data stream arriving at the node via one communication channels is transmitted from the node via other communication channels, taking into account the required transmission route.

A hub in a data transmission network is a device that combines the load of several data transmission channels for subsequent transmission over a smaller number of channels. The use of hubs allows you to reduce the cost of organizing communication channels that provide connection of subscribers to the data transmission network.

A communication channel is a set of technical means and a propagation medium that ensures the transmission of a message of any kind from a source to a recipient using telecommunication signals.

The structure of the computer network, built on the principle of organizing the exchange of information through the switching nodes of the data transmission network, assumes that network subscribers do not have direct (dedicated) communication channels among themselves, but connects to the nearest switching node and through it (and other intermediate nodes) with any other subscriber of this or even another computer network.

The advantages of building computer networks with the use of switching nodes of the data transmission network are: a significant reduction in the total number of communication channels and their length due to the absence of extremely important organization of direct channels between different network subscribers; a high degree of utilization of the bandwidth of communication channels due to the use of the same channels for transferring various types of information between network subscribers; the ability to unify technical solutions for software and hardware exchange for various network subscribers, including the creation of integrated service nodes capable of switching information flows containing data, voice, telefax and video signals.

Today, data transmission networks use three switching methods: circuit switching, message switching and packet switching.

When switching channels, a direct connection is created in the network by creating an end-to-end data transmission channel (without intermediate accumulation of information during transmission). The physical meaning of channel switching is, in fact, that before the start of information transmission in the network through the switching nodes, a direct electrical connection is established between the sending subscriber and the message recipient. Such a connection is established by sending a special call message by the sender, “contains the number (address) of the called subscriber” and, when passing through the network, occupies communication channels along the entire path of the subsequent message transmission. Obviously, when switching channels, all the constituent parts of the formed end-to-end communication channel must be free. In the event that in any part of the network the passage of the call will not be ensured (for example, no free channels between the switching nodes that make up the message transmission path), then caller receives a refusal to establish a connection and his call is considered lost for the network To send a message, the sending subscriber must repeat the call

After the connection is established, the sending subscriber receives a message that he can start data transfer. A fundamental feature of channel switching is that all channels occupied when establishing a connection are used in the process of data transfer simultaneously and are released only after the completion of data transfer between subscribers. A typical example of a circuit-switched network is a telephone network.

When messages are switched, the message is received and accumulated in the switching node, and then its subsequent transmission is carried out. This definition implies the main difference between message switching and circuit switching, ĸᴏᴛᴏᴩᴏᴇ consists essentially in the fact that when messages are switched, messages are intermediate stored in switching nodes and processed (definition of message priority, multiplication for multicast, message recording and archive, etc.). To process messages, they must have a format accepted in the network, that is, the same type of arrangement of individual message elements. The message from the subscriber first goes to the network switching node to which the subscriber is connected. Further, the node processes the message and determines the direction of its further transmission, taking into account the address. If all channels in the selected transmission direction are busy, the message waits in the queue for the moment when the desired channel is released. After the message reaches the network node to which the recipient subscriber is connected, the message is issued to him in full via the communication channel between this node and the subscriber. Τᴀᴋᴎᴍ ᴏϬᴩᴀᴈᴏᴍ, a message while passing through the network at any given time occupies only one communication channel.

Packet switching is defined as a type of message switching in which messages are split into pieces called packets and transmitted, received, and accumulated as such data packets.

These packets are numbered and provided with addresses, which allows them to be transmitted over the network simultaneously and independently of each other.

Nearly every modern company has a need to improve the efficiency of networks and computer systems technology. One of the prerequisites for this is the smooth transfer of information between servers, data stores, applications and users. It is the method of data transmission in information systems that often becomes a performance bottleneck, negating all the advantages of modern servers and storage systems. Developers and sysadmins try to remove the most obvious bottlenecks, even though they know that once a bottleneck is removed in one part of the system, it occurs in another.

Over the years, bottlenecks have arisen predominantly on servers, but as servers have evolved in a functional and technological manner, they have begun to migrate to networks and networked storage systems. Recently, very large storage arrays have been created, which brings bottlenecks back to the network. Data growth and centralization, as well as the bandwidth requirements of next-generation applications, often consume all available bandwidth.

When an information service manager is faced with the task of creating a new or expanding an existing information processing system, one of the most important questions for him will be the choice of data transmission technology. This problem includes not only the choice of network technology, but also the protocol for connecting various peripheral devices. The most popular storage area network (SAN) solutions are Fiber Channel, Ethernet, and InfiniBand.

Ethernet technology

Today, Ethernet technology is at the forefront of the high-performance LAN sector. Businesses all over the world are investing in Ethernet cabling and equipment and in staff training. The widespread use of this technology makes it possible to keep low prices on the market, and the cost of implementing each new generation of networks tends to decrease. The constant growth in traffic volume in modern networks is forcing operators, administrators and architects corporate networks look to faster network technologies to solve the problem of bandwidth shortages. Add to family Ethernet standard 10-Gigabit Ethernet enables new LAN applications to be supported.

Introduced over a quarter of a century ago, Ethernet technology soon dominated local area networks. Due to the simplicity of installation and maintenance, reliability and low cost of implementation, its popularity has grown so much that today we can safely say that almost all traffic on the Internet begins and ends in Ethernet networks. The IEEE 802.3ae 10-Gigabit Ethernet standard, approved in June 2002, marks a turning point in the development of this technology. With its introduction, Ethernet is expanding to include metropolitan (MAN) and wide area (WAN) networks.

There are a number of market factors that industry analysts say are driving 10-Gigabit Ethernet technology to the fore. In the development of network technologies, the emergence of an alliance of development companies has already become traditional, the main task of which is to promote new networks. 10-Gigabit Ethernet is no exception. At the origins of this technology was the 10-Gigabit Ethernet Alliance (10 GEA), which included such industry giants as 3Com, Cisco, Nortel, Intel, Sun and many other (more than a hundred) companies. While in previous versions of Fast Ethernet or Gigabit Ethernet, developers borrowed certain elements of other technologies, the specifications of the new standard were created practically from scratch. In addition, the 10-Gigabit Ethernet project was focused on large transport and backbone networks, for example, city-wide, while even Gigabit Ethernet was developed exclusively for use in local networks.

The 10-Gigabit Ethernet standard provides for the transmission of information streams at speeds up to 10 Gbps over single and multimode optical cable. Depending on the transmission medium, the distance can be between 65 m and 40 km. The new standard was supposed to meet the following basic technical requirements:

bidirectional data exchange in full duplex mode in point-to-point networks;
support for 10 Gbps data transfer rate at the MAC level;
specification of the physical layer LAN PHY for connecting to local networks, operating at the MAC layer with a data transfer rate of 10 Gbit / s;
WAN PHY physical layer specification for SONET / SDH connectivity, operating at the MAC layer with a data rate compliant with the OC-192 standard;
determining a mechanism for adapting the data rate of the MAC layer to the data rate of the WAN PHY;
support for two types of fiber optic cable - single mode (SMF) and multimode (MMF);
XGMII * Media Independent Interface Specification;
backward compatible with previous versions Ethernet (saving packet format, size, etc.).

* XG here stands for 10 Gigabit and MII stands for Media Independent Interface.

Recall that the 10/100 Ethernet standard defines two modes: half duplex and full duplex. Half-duplex in the classic version provides for the use of a shared transmission medium and the CSMA / CD (Carrier-Sense Multiple Access / Collision Detection) protocol. The main disadvantages of this mode are the loss of efficiency with an increase in the number of simultaneously operating stations and distance restrictions associated with the minimum packet length (64 bytes). Gigabit Ethernet uses a carrier expansion technique to keep the packet length to a minimum, which expands it to 512 bytes. Because 10-Gigabit Ethernet is focused on point-to-point backbones, half-duplex is not part of the specification. Therefore, in this case, the channel length is limited only by the characteristics of the physical medium used by the transmitting / receiving devices, signal power and modulation methods. The required topology can be provided, for example, using switches. Full duplex transmission also allows a minimum burst size of 64 bytes to be maintained without the use of carrier extension techniques.

In accordance with the reference model for open systems interconnection (OSI), network technology is defined by two lower layers: physical (Layer 1, Physical) and channel (Layer 2, Data Link). In this scheme, the layer of physical devices Ethernet (PHY) corresponds to Layer 1, and the layer of media access control (MAC) - Layer 2. In turn, each of these layers, depending on the technology being implemented, can contain several sublayers.

The Media Access Control (MAC) layer provides a logical connection between MAC clients of peer-to-peer (peer-to-peer) workstations. Its main functions are to initialize, manage and maintain a peer-to-peer connection. Obviously, the normal data transfer rate from the MAC layer to the PHY for 10 Gigabit Ethernet is 10 Gbps. However, the WAN PHY layer must transmit data at a slightly lower rate to accommodate SONET OC-192 networks. This is achieved using a mechanism for dynamic adaptation of the interframe interval, which provides for its increase by a predetermined period of time.

The Reconciliation Sublayer (Figure 1) is the interface between the serial data stream of the MAC layer and the parallel stream of the XGMII sublayer. It maps octets of MAC layer data to parallel XGMII paths. XGMII is a 10 Gigabit environment independent interface. Its main function is to provide a simple and easily implemented interface between the link and physical layers. It isolates the link layer from the specifics of the physical and thereby allows the former to work at a single logical level with different implementations of the latter. The XGMII consists of two independent transmit and receive channels, each carrying 32 bits of data over four 8-bit paths.

Rice. 1. Layers of 10-Gigabit Ethernet.

The next part of the protocol stack is related to the physical layer of 10 Gigabit Ethernet. The Ethernet architecture divides the physical layer into three sub-layers. The Physical Coding Sublayer (PCS) encodes / decodes the data stream coming from and to the link layer. The Physical Media Attachment (PMA) sublayer is a parallel-to-serial (forward and reverse) converter. It performs the conversion of a group of codes to a bitstream for serial bit oriented transmission and inverse conversion. The same sublayer provides transmit / receive synchronization. The Physical Media Dependent (PMD) sublayer is responsible for signaling in a given physical medium. Typical functions of this sublevel are signal shaping and amplification, modulation. Different PMDs support different physical media. In turn, the Media Dependent Interface (MDI) defines the connector types for different physical media and PMD devices.

10-Gigabit Ethernet technology provides low cost of ownership compared to alternatives, including both acquisition and maintenance costs, as customers' existing Ethernet infrastructure is easily interoperable with it. In addition, 10 Gigabit Ethernet appeals to administrators with a familiar management organization and the ability to leverage lessons learned as it leverages the processes, protocols and controls already deployed in the existing infrastructure. It is worth recalling that this standard provides flexibility in the design of connections between servers, switches, and routers. Thus, Ethernet technology offers three main benefits:

ease of use,
high throughput,
low cost.

In addition, it is simpler than some other technologies, because it allows you to link networks located in different places, as part of a single network. Ethernet bandwidth is scalable in increments of 1 Gbps to 10 Gbps to make better use of network capacity. Finally, Ethernet equipment is generally more cost effective than traditional telecommunications equipment.

To illustrate the capabilities of the technology, we will give one example. Using a 10-Gigabit Ethernet network, a team of scientists working on the Japanese Data Reservoir project (http://data-reservoir.adm.su-tokyo.ac.jp) transmitted data from Tokyo to the Elementary Physics Research Center in Geneva. particles CERN. The data line crossed 17 time zones and was 11,495 miles (18,495 km) long. A 10-Gigabit Ethernet line connected computers in Tokyo and Geneva as part of the same local area network. The network used optical equipment and Ethernet switches from Cisco Systems, Foundry Networks and Nortel Networks.

In recent years, Ethernet has also become widely used by telecom operators - to connect objects within the city. But Ethernet can stretch even further, spanning entire continents.

Fiber Channel

Fiber Channel technology makes it possible to fundamentally change the architecture of a computer network for any large organization. The fact is that it is well suited for the implementation of a centralized SAN storage system, where disk and tape drives are located in their own separate network, including geographically quite far from the main corporate servers. Fiber Channel is a serial link standard designed for high-speed communications between servers, storage, workstations and hubs and switches. Note that this interface is almost universal; it is used not only for connecting individual drives and data storages.

When the first networks appeared to bring computers together for collaboration, it was convenient and effective to bring resources closer to workgroups. Thus, in an attempt to minimize the network load, the storage media were evenly split across multiple servers and desktops. There are two data transmission channels simultaneously in the network: the network itself, through which there is an exchange between clients and servers, and the channel through which data is exchanged between system bus computer and storage device. This can be a link between the controller and the hard drive, or between the RAID controller and an external disk array.

This separation of channels is largely due to the different requirements for data transfer. In the network, the first place is the delivery of the necessary information to one client out of many possible ones, for which it is necessary to create certain and very complex addressing mechanisms. In addition, the network channel requires significant distances, so a serial connection is preferred here for data transmission. But the storage channel performs an extremely simple task, providing the ability to exchange with a previously known data storage device. The only thing that is required of him is to do it as quickly as possible. Distances are usually short here.

However, today's networks are faced with the challenges of processing more and more data. High-speed multimedia applications, image processing require much higher I / O speed than ever before. Organizations are forced to store more and more data online, which requires more external storage capacity. The need for insurance copying of huge amounts of data requires the separation of secondary storage devices at ever greater distances from the processing servers. In some cases, it turns out that pooling server and storage resources into a single pool for a data center using Fiber Channel is much more efficient than using a standard set of Ethernet plus SCSI.

The ANSI Institute registered a working group for the development of a method for high-speed data exchange between supercomputers, workstations, PCs, storage devices and display devices back in 1988. And in 1992, the three largest computer companies - IBM (http://www.ibm.com ), Sun Microsystems (http://www.sun.com) and HP (http://www.hp.com) formed the Fiber Channel Systems Initiative (FSCI) tasked with developing a method for fast digital data transfer ... The group has developed a number of preliminary specifications - profiles. Since fiber-optic cables were to become the physical medium for the exchange of information, the word fiber appeared in the name of the technology. However, a few years later, the ability to use copper wires was added to the corresponding recommendations. Then the ISO committee (International Standard Organization) proposed replacing the English spelling of fiber with the French fiber, in order to somehow reduce the association with the fiber-optic environment, while preserving almost the original spelling. When the preliminary work on the profiles has been completed, further work on support and development new technology was taken over by the Fiber Channel Association (FCA), which became an organizational member of the ANSI committee. In addition to the FCA, an independent working group FCLC (Fiber Channel Loop Community), which began to promote one of the variants of Fiber Channel technology - FC-AL (Fiber Channel Arbitrated Loop). Currently, the FCIA (Fiber Channel Industry Association, http://www.fibrechannel.org) has taken over all the coordination work to promote Fiber Channel technology. In 1994, the FC-PH (physical connection and data transfer protocol) standard was approved by the ANSI T11 committee and received the designation X3.203-1994.

Fiber Channel technology has a number of advantages that make this standard convenient for organizing data exchange among groups of computers, as well as when used as an interface to mass storage devices, in local networks and when choosing a means of access to WANs. One of the main advantages of this technology is its high data transfer rate.

FC-AL is just one of three possible Fiber Channel topologies used in storage, in particular. In addition, point-to-point topology and star topology based on switches and hubs are possible. A network that is built on the basis of switches connecting many nodes (Fig. 2) is called a fabric in Fiber Channel terminology.

Rice. 2. Factory based on Fiber Channel.

Up to 126 hot-swappable devices can be connected to the FC-AL loop. When using a coaxial cable, the distance between them can reach 30 m, while in the case of a fiber-optic cable, it increases to 10 km. The technology is based on the technique of simply moving data from the transmitter buffer to the receiver buffer with full control over this and only this operation. For FC-AL, it does not matter at all how the data is processed by individual protocols before and after being placed in the buffer, so the type of data transmitted (commands, packets or frames) does not matter.

The Fiber Channel architectural model describes in detail the connection parameters and communication protocols between individual nodes. This model can be represented as five functional layers that define the physical interface, transmission protocol, signaling protocol, general procedures, and display protocol. The numbering goes from the lowest hardware level FC-0, which is responsible for the parameters of the physical connection, to the top software FC-4, which interacts with applications more high level... The display protocol provides communication with I / O interfaces (SCSI, IPI, HIPPI, ESCON) and network protocols (802.2, IP). In this case, all supported protocols can be used simultaneously. For example, the FC-AL interface, which works with IP and SCSI protocols, is suitable for both system-to-system and system-to-peripheral exchange. This eliminates the need for additional I / O controllers, significantly reducing cabling complexity and of course overall cost.

Since Fiber Channel is a low-level protocol that does not contain I / O commands, communication with external devices and computers is provided by higher-level protocols such as SCSI and IP, for which FC-PH serves as a transport. Network and I / O protocols (such as SCSI commands) are converted to FC-PH frames and delivered to the destination. Any device (computer, server, printer, storage) that can communicate using Fiber Channel technology is called a Node port, or simply a node. Thus, the main purpose of Fiber Channel is the ability to manipulate high-level protocols using different transmission media and pre-existing cabling systems.

High reliability of exchange when using Fiber Channel is due to the dual-port architecture of disk devices, cyclic control of transmitted information and hot-swappable devices. The protocol supports almost any cable system in use today. However, the most widespread are two carriers - optics and twisted pair. Optical links are used to connect between devices on a Fiber Channel network, while twisted-pair cables are used to connect individual components in a device (for example, disks in a disk subsystem).

The standard provides for several bandwidths and provides an exchange rate of 1, 2, or 4 Gb / s. Taking into account the fact that two optical cables are used to connect devices, each of which works in the same direction, with a balanced set of read-write operations, the data exchange rate doubles. In other words, Fiber Channel operates in full duplex mode. In terms of megabytes, the rated speed of Fiber Channel is 100, 200 and 400 MB / s, respectively. In reality, with a 50% ratio of read-write operations, the interface speed reaches 200, 400 and 800 MB / s. Currently, 2 Gbps Fiber Channel solutions are the most popular because they offer the best price / performance ratio.

Note that Fiber Channel equipment can be roughly divided into four main categories: adapters, hubs, switches and routers, and the latter have not yet become widespread.

Fiber Channel solutions are typically designed for organizations that need to maintain large volumes of information online, speed up primary and secondary storage for data-intensive networks, and remove storage from servers over longer distances. allowed in the SCSI standard. Typical applications for Fiber Channel solutions are databases and databanks, analysis and decision support systems based on large amounts of data, storage and processing systems for multimedia information for television, film studios, as well as systems where disks must be remote from servers. for security reasons.

Fiber Channel makes it possible to separate all data streams between enterprise servers, data archiving, etc. from the user's local network. In this option, the configuration possibilities are enormous - any server can access any disk resource allowed by the system administrator, it is possible to access the same disk for several devices simultaneously, and at a very high speed. This option also makes data archiving an easy and transparent task. At any time, you can create a cluster, freeing up resources for it on any of the Fiber Channel storage systems. Scaling is also quite clear and understandable - depending on what capabilities are missing, you can add either a server (which will be bought based solely on its computing capabilities) or a new storage system.

One of the very important and necessary features of Fiber Channel is the ability to segment or, as they say, zoning the system. Zones are similar to Virtual LANs on a local area network — devices in different zones cannot "see" each other. Zoning is possible either through Switched Fabric or WWN (World Wide Name). The WWN address is similar to the MAC address in Ethernet networks, each FC controller has its own unique WWN address, which is assigned by the manufacturer, and any correct storage system allows you to enter the addresses of those controllers or matrix ports that this device is allowed to work with. Zoning is primarily intended to improve the security and performance of SANs. Unlike a regular network, it is impossible to gain access to a device closed for a given zone from the outside world.

FICON technology

FICON (FIber CONnection) technology provides increased performance, extended functionality and communication over long distances. As a data transfer protocol, it is based on the ANSI Fiber Channel (FC-SB-2) standard. IBM's first general-purpose standard for communication between mainframes and external devices (such as disks, printers, and tape drives) was based on parallel connections, not too different from the multicore cables and multi-pin connectors that were used in those years to connect desktop printers to PCs. ... Many parallel wires were used to carry more data "at a time" (in parallel); in mainframes it was called bus and tag.

Physically oversized connectors and cabling were the only way to communicate until they hit the market in the 1990s. ESCON technology. It was a fundamentally different technology: for the first time, instead of copper, optical fiber was used and data was transmitted not in parallel, but in series. Everyone was well aware that ESCON was much better and significantly faster, at least on paper, but before the general adoption of the technology, it took a lot of testing and effort to convince buyers. The ESCON technology is believed to have emerged during a stagnant market; in addition, devices supporting this standard were presented with a noticeable delay, so the technology met with a cool reception, and it took almost four years for its widespread adoption.

With FICON, history has largely repeated itself. For the first time this technology was presented by IBM on S / 390 servers back in 1997. It was immediately clear to many analysts that this is in many ways a technically more advanced solution. However, for several years, FICON has been used almost exclusively for connecting tape drives (a significantly improved solution for the purpose of creating backups and recovery) and printers. It wasn't until 2001 that IBM finally equipped FICON with its Enterprise Storage Server, codenamed Shark. This event again coincided with a severe economic downturn, when the introduction of new technologies in enterprises slowed down. Literally a year later, a number of circumstances arose that contributed to the accelerated adoption of FICON. This time around, the concept of fiber was no longer new, and storage area network (SAN) technologies were widespread both in the mainframe world and beyond.

The storage market continues to grow steadily. Today's devices called directors, originally designed to support ESCON, now support Fiber Channel and deploy FICON solutions on these same devices. According to the developers, FICON provides significantly more functionality than Fiber Channel.

InfiniBand

The InfiniBand architecture defines a common standard for handling communications, networking, and storage I / O. This new standard led to the formation of the InfiniBand Trade Association (IBTA, http://www.infinibandta.org). Simply put, InfiniBand is a next-generation I / O architecture standard that takes a networked approach to connecting data center servers, storage systems, and networking devices.

InfiniBand was designed as an open solution that could replace all other networking technologies in a wide variety of areas. This also applied to common LAN technologies (all types of Ethernet and storage networks, in particular Fiber Channel), and specialized cluster networks (Myrinet, SCI, etc.), and even connecting I / O devices to a PC as a possible replacement PCI buses and I / O channels such as SCSI. In addition, the InfiniBand infrastructure could serve to consolidate chunks using different technologies into a single system. The advantage of InfiniBand over specialized, high-performance cluster-oriented networking technologies lies in its versatility. Oracle, for example, supports InfiniBand in its clustering solutions. A year ago, HP and Oracle set a TPC-H performance record (for 1TB databases) on a ProLiant DL585-based InfiniBand cluster running Oracle 10g on Linux. In the summer of 2005, IBM achieved record highs for TPC-H (for 3TB databases) in a DB2 and SuSE Linux Enterprise Server 9 environment in an xSeries 346-based InfiniBand cluster. At the same time, the achieved cost per transaction was almost half of from the closest competitors.

Using a technique called a switched network fabric, or lattice, InfiniBand shifts I / O traffic from server processors to peripherals and other processors or servers throughout the enterprise. A special cable (link) is used as a physical channel, providing a data transfer rate of 2.5 Gbps in both directions (InfiniBand 1x). The architecture is organized as a multi-tier architecture with four hardware layers and upper layers implemented in software. In each physical channel, you can organize many virtual channels, assigning them different priorities. To increase the speed, there are 4x and 12x versions of InfiniBand, which use 16 and 48 wires, respectively, and the data transfer rates over them are 10 Gbps (InfiniBand 4x) and 30 Gbps (InfiniBand 12x).

Solutions based on the InfiniBand architecture are in demand in four main markets: corporate data centers (including data warehouses), high-performance computer clusters, embedded applications and communications. InfiniBand technology allows industry-standard servers to be clustered to provide datacenter performance, scalability, and resilience — capabilities typically found only in high-end platforms worth millions of dollars. In addition, InfiniBand storage can be connected to server clusters, which allows all storage resources to be linked directly to compute resources. The high-performance cluster market is aggressively seeking new ways to expand computational capabilities and can therefore greatly benefit from the high throughput, low latency, and superior scalability offered by low-cost InfiniBand products. Embedded applications such as military systems, real-time systems, video streaming, and more will greatly benefit from the reliability and flexibility of InfiniBand connections. In addition, the communications market is constantly demanding increased bandwidth, which is achieved with 10Gbps and 30Gbps InfiniBand connections.

At the physical layer of the InfiniBand protocol, electrical and mechanical characteristics are defined, including fiber and copper cables, connectors, parameters that define the properties of hot swapping. At the level of links, the parameters of transmitted packets, operations connecting point to point, and switching features in the local subnetwork are defined. The network layer defines the rules for routing packets between subnets; within a subnet, this layer is not required. The transport layer provides packet assembly into a message, channel multiplexing, and transport services.

Let's note some key features InfiniBand architecture. I / O and clustering uses a single InfiniBand card in the server, eliminating the need for separate communication and storage cards (however, for a typical server, it is recommended to install two of these cards configured for redundancy). Just one connection to the InfiniBand switch is enough for each server, IP network or SAN system (redundancy is reduced to simple duplication of the connection to another switch). Finally, the InfiniBand architecture resolves connectivity and bandwidth constraints within the server while still providing the required bandwidth and communications capability for external systems storage.

The InfiniBand architecture consists of the following three main components (Figure 3). HCA (Host Channel Adapter) is installed inside a server or workstation that acts as a master (host). It acts as an interface between the memory controller and the outside world and serves to connect host machines to the network infrastructure based on InfiniBand technology. The HCA implements the messaging protocol and the underlying mechanism direct access to memory. It connects to one or more InfiniBand switches and can exchange messages with one or more TCAs. Target Channel Adapter (TCA) is designed to connect devices such as drives, disk arrays, or network controllers to the InfiniBand network. It, in turn, serves as the interface between the InfiniBand switch and the peripheral I / O controllers. These controllers do not have to be of the same type or belong to the same class, which allows different devices to be combined into one system. Thus, the TCA acts as a physical middleware between the data traffic of the InfiniBand fabric and the more traditional I / O controllers for other subsystems such as Ethernet, SCSI, and Fiber Channel. It should be noted that TCA can interact with HCA directly. InfiniBand switches and routers provide central docking points, and multiple TCAs can be connected to the management HCA. InfiniBand switches form the core of the network infrastructure. With the help of multiple channels, they are connected to each other and to the TCA; in this case, mechanisms such as channel grouping and load balancing can be implemented. If the switches operate within a single subnet formed by directly connected devices, then InfiniBand routers combine these subnets, establishing communication between multiple switches.

Rice. 3. The main components of the SAN-network based on InfiniBand.

Most of the advanced logic capabilities of the InfiniBand system are built into the adapters that connect the nodes to the I / O system. Each adapter type offloads the host from performing transport tasks by using an InfiniBand link adapter, which is responsible for organizing I / O messages into packets to deliver data over the network. As a result, the OS on the host and the server processor are freed from this task. It is worth noting that such an organization is fundamentally different from what happens in communications based on the TCP / IP protocol.

InfiniBand defines a highly flexible set of communications links and transport layer mechanisms to fine tune InfiniBand SAN characteristics based on application requirements, including:

packages of variable size;
maximum size of a transmission unit: 256, 512 bytes, 1, 2, 4 KB;
Layer 2 Local Route Headers (LRH) to route packets to the correct port on the channel adapter;
additional layer 3 header for global routing (GRH, Global Route Header);
multicast support;
variant and invariant checksums (VCRC and ICRC) to ensure data integrity.

The maximum transmission unit size determines system characteristics such as packet timing jitter, encapsulation overhead, and latency used when designing systems with multiple protocols. The ability to omit the global route information when forwarding to a local subnet destination reduces local communication overhead. The VCRC code is recalculated every time the next link of the communication channel passes, and the ICRC code is calculated when the packet is received by the destination, which guarantees the integrity of the transmission along the link and along the entire communication channel.

InfiniBand defines permission-based flow control — to prevent head of line blocking and packet loss — as well as link-layer flow control and end-to-end flow control. Permission-based link layer control is superior to the widely used XON / XOFF protocol, eliminating maximum range restrictions and providing better link utilization. The receiving end of the communication line sends permission to the transmitting device indicating the amount of data that can be received reliably. No data is transmitted until the receiver sends a permission indicating there is free space in the receive buffer. A mechanism for transferring permissions between devices is built into the protocols of connections and communication lines to ensure reliable flow control. Link layer flow control is organized on a per VC basis to prevent transmission collisions common to other technologies.

With InfiniBand, communication with remote storage modules, networking and server-to-server connections will be accomplished by connecting all devices through a central, unified switch and link structure. The InfiniBand architecture allows I / O devices to be placed up to 17 m from the server using copper wire, up to 300 m using multimode fiber optic cable, and up to 10 km using single mode fiber.

Today, InfiniBand is slowly gaining popularity again as a backbone technology for server and storage clusters, and in data centers as the basis for connections between servers and storage systems. An organization called the OpenIB Alliance (Open InfiniBand Alliance, http://www.openib.org) is doing a lot in this direction. Specifically, the alliance aims to develop a standard open source InfiniBand software support stack for Linux and Windows. A year ago, support for InfiniBand technology was officially included in the Linux kernel. In addition, at the end of 2005, OpenIB representatives demonstrated the possibility of using InfiniBand technology over long distances. The best achievement during the demo was data transmission at 10 Gbps over a distance of 80.5 km. The experiment involved data centers of a number of companies and scientific organizations. At each endpoint, InfiniBand was encapsulated on SONET OC-192c, ATM, or 10 Gigabit Ethernet interfaces without sacrificing bandwidth.