Data 20 Communication 20basic 20 Concepts
Data 20 Communication 20basic 20 Concepts
Data 20 Communication 20basic 20 Concepts
Concepts
Data Communications:
Basic Concepts
Datapro Summary
Data communications is an integral part of business. Whether a data network accommodates 10 personal
computers on a LAN or 100 nodes in a global network, data communications is the link for greater productivity,
efficiency. and cost savings. This report offers an overview of data communications principles, the products and
services comprising a data network, and the issues facing this dynamic industry.
43~ Thomas No//e
Pwsid& C/MI Corp.
Lipdated by Datapro staff
..
l Both agreed on the strategyw to be used to transfer the information and the structure of the information itself.
Data communications can be viewed as the set of strategies needed to ensure these conditions.
with an error-checking sequence designed to help the receiver detect block errors. But because blocks are likely
to be long, the transmitter and receiver cannot track accurately and might lose count of where characters begin
and end (“lose svnc”).
d To prevent this, synchronous channels provide both sender and receiver a standard clock
signal.
Asynchronous communications is inefficient when used with devices that can coHect and configure
information into blocks, but it is still very common for simple connections where the characters sent on a line
are keyed by, or displayed to, a human operator. Synchronous data Iinks are best implemented where direct
human intervention is not possible because of the speed. The ability of a human to view the result of a
connection means that little must be done by the computer!terminaI to protect data; the “protocol” is that there is
no special procedure. Thus. the term “asynchronous protocol” means that characters are sent as they are keyed or
as they are to be displayed, with little or no control dialog.
Synchronous blocks. having a potential strategy for block error detection and correction, can justify a more
complex set of rules for information exchange. “Synchronous protocols” are therefore more complicated.
employing strategies for detecting and correcting errors, controlling the rate of flow, and setting other
characteristics of connections.
Communications Standards
The need for both parties in a connection to agree on data presentation and dialog controt rules, the protocol, has
already been noted. Standards to define these rules are as old as data communications and arise from three major
sources:
l Vendors thetnsetves whose proprietary protocols may be “open” to support by other vendors because the
originating vendor publishes the specifications. IBM’s Systems Network Architecture (SNA) is an example of a
proprietary, but “open” protocol.
l Trade groups or consortiums, which represent special interests in a given communications market. The
Institute of Electrical and Electronic Engineers (IEEE) is a trade group that has protnoted the basic standards for
local area networking, IEEE 802.
l National or international standards bodies, which formally debate rules and publish standards. In the U.S., the
American National Standards Institute (ANSI) and the National Institute for Standards and Technology (NIST,
formerly the National Bureau of Standards) are the principal standards bodies. The International Organization
for Standardization (ISO) and the International TeIecotnmunications Union-Telecommunications
Standardization Sector (KU-TSS. formerly known as the CCITT) are the two international standards groups
most involved in data comtnunications.
The OSI model is not itself a standard, but a framework wlhich describes the relationship of standards. There are
standards for each of the seven OSI model layers, often several at the same layer.
Data communications and data networks serve the information processing systems used, or contemplated, by a
business. All information technology planning is based on estabt ishing a user-to-source relationship and
identifying the technology elements needed to support it. Data networks are such an element, and data
cotnmunications is the foundation of data networks.
Early data communications applications were developed when computer facilities were so expensive that
access to a cotnputer had to be given to numerous users via relatively primitive entry/display terminals. The
computer processed and fortnatted all information at its central location. But as technology advanced, it created
the microprocessor and enabled the development of inexpensive desktop computer systems. These allowed
cotnputing power and information storage to be dispersed. In this new environment, four distinct types of data
cotntnunications relationships developed: host/terminal, client/server, peer or distributed processing, and
internetworking. Figure “C(~nlnlz(17icatior~sNetwork Relationships and Topologies” shows an example of the way that
each type of relationship affects the structure of the network that must support it.
HoWTerminal Relationships
Normallv, host/terminal applications occur when the process of information entry and display is the major
element bf the application, and the goal is to support the fastest rate of acquisition or output of data. In almost all
cases, the system interaction is to a human operator, either directly (via keyboard and display) or indirectly (via
printout).
In host/terminal applications, the speed of the human/mechanical component of the connection is often low
enough to limit the information flow rate to a level well below that of channel capacity. Because of this,
host/terminal networks often include facilities to share the information channel among multiple terminals to
reduce overall cost. These devices are called “terminal servers” or “cluster controllers.” IBM’s popular 3270
family of devices includes the 3 174 cluster controller.
When a desktop computer is used to “emulate” a terminal, the interaction between the personal or other
desktop computer and its partner system is still considered a host/terminal interaction. Any processing
capabilities of the desktop system are “hidden” by the fact that the PC is emulating a dumb terminal.
Host/tertninaI applications are forgiving of channel limitations. Their relatively limited speed has already been
noted: high-capacity channels can be justified only by sharing them among multiple terminals. Host/terminal
applications are also generally immune to delays in the data path, since the human reaction titne is normally tong
enough to hide any network transit delay.
Networks for support of host/tertninal relationships take on a “tree” structure, as shown in Figure
“Comntrnicatiom Netwwk Relationships and Topologies. ” The tertninals are often concentrated via cluster controllers
or servers onto a shared trunk, which may be further concentrated to a higher-speed facility. All infortnation
paths lead to the computer systetn at the heart of the structure, and there is no connection between users except
through that computer.
Client/Server Relationships
Client/server applications utilize a small computer at the point of human/system interaction and a larger one as a
central repository for infortnation and/or information processing power. The client system can provide local
services to its user, but it may from time to time require access to information stored at the server or to the
server’s specialized processing resources. When this happens, the client reties on a data communications
connection to the server. The file sharing and printer sharing done in PC-based local area networks (LANs) is a
common example of a pritnitive fortn of client/server computing.
Because client/server applications are between computing systems and not between a system and a human, the
speed of the exchange of infortnation is not limited to human rates. Thus, the applications utilize a much higher
channel capacity for the brief period of the interaction, though the capacity might be wasted during periods when
the client svstem was involved in a local user dialog only, w Client/server applications, therefore, benefit from
strategies for sharing infortnation channels as well.
The extent to which the client and server systetns interact in satisfying a user need varies considerable. Some
systems, such as electronic tnail systems, simply deliver a message to a client to be read by its user at ati
convenient titne. In this case, the client/server interaction is relatively infrequent and not highly constrained by
performance. But if the client system is processing a remote database, each record may be sent over the network.
In the latter case, network performance will have a major impact on the application. Client/server relationships
often place information sources farther out toward the network user? often at the points of concentration. To
provide access to these resources without loading the host, intermediate cross-connections are often provided,
creating a meshing of these concentration points.
Peer Relationships
A client/server application requires two smart system devices, for example, a desktop computer and a data
center system. However, despite the fact that both devices are computers, there is a master/slave relationship
inherent in the fact that the server is a source of information--often for many clients.
Peer applications have no such inherent master. Peer systems are those that have relatively little difference in
information storage or processing capacitv and are capable of adopting virtually any sort of relationship with
one another according to the momentary ieeds of the application or user.
Peer connectivity is the most challenging of all types of connectivity to provide, since there is no preferred
information focus among the systems communicating. Without such a focus, any number of connections and
flow volumes could be possible, and the capacity and number of channels needed to support them make design
of a total network difficult. The unpredictability also makes it difficult to concentrate traffic for efficient use of
circuits; there are no consistent partners to create consistent patterns of flow. A peer network, therefore, tends to
connect users at all levels.
There are few/ true peer applications today, because most companies have central data center resources or other
departmental information storage points. Peer networking is most likely to be found in companies that rely on
personal computers or desktop UNIX systems.
H
Internetworking Relationships
All of the relationships described so far have been between information systems and have been explained in an
application context. The last relationship, internetworking, is not a system relationship at all. but a network
relationship affecting all users on the network.
Internetworking is most likely to occur when a business that has previouslv planned data communications on a
per-application basis begins to consider it as a kev part of its strategic planning. A large part of creating a
strategic network is making information access Miithin the firm more universal, something that is often called
“building an enterprise network.” reflecting the breakdown of internal network barriers.
In a technical sense, inter-networking is the task of building a single, large network by combining existing ones
while retaining the application support characteristics of each of the networks. Figure “Communications Net-work
Relationships and Topologies” shows an internetwork structure created by linking a series of LANs. It is an area
supported by specialized products discussed in the Switching Devices section later in this report.
or higher. They provide the benefits of wideband digital service at a lower cost than full Tl/El.
Switched digital carrier services provide the bandwidth of a fractional or full TlfEI WAN link through
l
multiple, independent 56K/64K bps dial-up connections. Multiplexing equipment owned by the user establishes
these dial-up links through a switched digital network and evenly distributes the data, voice, or video
transmission across those links. Identical equipment at the destination, receiving those multiple transmissions,
resynchronizes and recombines them to restore the original message.
l The Integrated Services Digital Network (ISDN) is an ambitious plan to create a fully integrated switched
digital network. Basic Rate ISDN, the most economical form, offers two 64K bps user channels and one 16K
bps “signaling” channel. Primary Rate ISDN. a higher-capacity service, provides 23 (in the US.) or 3 I (in
Europe) user channels and I signaling channel, all operating at 64K bps.
A more detailed description of carrier services associated with data transmission, including ISDN and fractional
Tl, can be found in other Datapro reports.
Value-Added Services
All of the services described so far provide the user with unformatted transmission capacitv and can be used
with any type of information source/user, including fax, voice. or video. whose demand is Within the capacity of
the service. This generality of capabilitvw limits the capability of the service to offer specialized benefits to the
data user. A network designed onlvr’for data can optimize the transmission of data and reduce overall network
costs.
A Rand Corp. study of the 1960s showed that data can be broken into small “packets” of information and
moved through a network of shared trunks and “nodes” to its destination. The sharing of circuits and equipment
can result in a lower per-character cost than could be achieved through the use of “ban dwidth” services like
fractional Tl /E I. Th is study became the basis of vfalue-added, packet s witched data ne tworks.
Public packet network services differ from the dial-up or leased bandwidth services described earlier in several
important ways:
l There is a specific access protocol required to attach to the network and transfer information. Traditional
carrier services are protocol independent.
l Network services are priced based on usage, meaning the number of characters transmitted from source to
destination. The distance between users and the duration of the connection are not normally bit1 ing factors.
l The introduction of multiple shared trunks and nodes generates an appreciable delay in information transport,
often greater than that generated by a satellite data path. This can impact the performance of some applications.
Public packet data services are most useful when an application involves the support of a widely dispersed
population of “occasional” users.
Packet technology can also be employed in private networks. Because packet network interfaces are based on
international standards, such networks are excellent for interconnecting computers from different vendors.
Packet standards also form the foundation of the OS1 protocols, discussed briefly earlier in this report.
A recent industry development is fast-packet switching--a streamlined approach to packet processing
providing greater efficiency and lower transit delay for interactive data, voice/video, and multimedia
communications. Fast-packet switching assumes that the wide area connection. utilizing fiber optics rather than
copper, is virtual ly error-free. It elimi nates error checking, th erefore, at all but the destination node of a
transmission. On e type of fast -packet service, frame relay, is ideal for interactive LAN/WAN applications that
cannot tolerate delay. It propagates data in variable-length frames across star and mesh networks of any size.
Fractional and full Tl /E I frame relay products and services are widely available.
Asynchronous transfer mode (ATM) products and services, operating at T3/E3 speeds and higher, process
videoconferencing, and multimedia network transmissions in small fixed-length (53.byte) cells. thereby
minimizing delay and congestion. ATM switches and carrier service offerings have been emerging rapidly since
1994.
example).
Most data communications environments are much more complex. The need to utilize concentration to spread
the cost of carrier services among multiple users has already been noted. Wideband and broadband services not
only have capacity in excess of what most single applications can justify. they often have service interfaces that
most computers and terminals cannot directly support.
For these reasons, data communications equipment is most often really data networking equipment. Its
purpose is to establish a shared set of facilities that users and information sources can use to make connections
and to maintain those facilities in correct operation.
Interface Devices
Most transmission services, and all public carrier services, require some form of interface through which
computers and terminals can attach. The interface provides whatever transformation is required between the
digital interface on the computer or terminal and the carrier service itself. It also generally provides some status
indicators that enable the computer or terminal to determine if the service (and the interface device) is operating
properly. These are called “control signals.” in that they provide for the local control of the interface.
Modems are the most common service interface, designed to link digital computers and terminals to each other
via analog carrier services. Modems can be classified as either synchronous (supporting the generation of the
synchronous clock signal) or asynchronous, and by the data transfer rates they support. Modems normally
transmit data at 2400, 9600, 14.4K, l9.2K, or 28K bps.
CSUs/DSUs are devices that interface communicating equipment to digital carrier services. Lower-speed
CSU/DSUs, designed to support narrowband digital services, are much like modems in appearance. The
higher-speed CSU/DSU devices, designed for wideband Tl, for example, are normally built into networking
devices that use the wideband interface.
Both modems and CSU/DSU devices can provide special services, going beyond simple interfacing. Data
compression uses one of several algorithms to contract the data stream generated by a data device, thereby
decreasing the number of characters being transmitted and increasing the effective throughput. Compression
rates of 2: I or 3: I are typical, and some information can be compressed even more. Network management
support on interface devices allows a user to monitor the quality of the circuit connecting the devices and to run
basic tests as well. Encryption prevents interception of information by encoding it. Backup features allow a
leased-line modem or CSU/DSU to dial a backup connection should the leased service fail.
Concentration Devices
Human-operated terminals are rarely utilized at a rate that taxes even a low-speed transmission facility, yet the
minimum dial or leased carrier circuit has a theoretical capacity of 2400 bps or more. Even for voice grade and
narrowband services, some form of concentration of multiple terminals onto a single circuit will improve
economy. Terminal servers and cluster controllers are devices that are provided by the computer vendor to
accomplish this concentration, but there are other devices as well.
Multiplexers are the most common form of third-party concentration device. There are several types of
multiplexers:
l Frequency-division multiplexers--separate multiple conversations by allocating each a different carrier
frequency to be modulated by the digital data. This type of multiplexing is old and extremely rare outside carrier
applications.
l Time-division multiplexer+-separate multiple conversations by allocating each a reserved amount of space in a
digital “frame” of data, which is transmitted at a regular interval. A TI or El frame consists of 24 or 32 eight-bit
bytes plus a framing bit, so Tl/E I is a form of time-division multiplexing.
l Statistical time-division multiplexers--identify information generated by each conversation through a “heade r”
code that prefixes the data. Capacity is allocated to a conversation when that conversation needs it.
0 Networking multipiexers--high-end statistical multiplexers performing any combination of concentration,
adaptation, and intelligent routing. Usually supporting several TI /E I aggregates, they enable multiple termina
devices and host computers to exchange information over star or mesh networks.
l Access tnultiplexers--provide access to an enterprise backbone for a remote office environment. Access
multiplexers usually concentrate data from multiple devices onto single or dual trunk connections leading to the
Switching Devices
The communications processor, usually designed for one or more vendor-specific families of computing and
communications equipment, performs iletwork control. intelligent routing, and concentration functions. As a
front end to a host computer, it serves as a master processor. relieving the host of the overhead involved in
message handling and network control. As an intelligent switch, it routes messages across the network. either
under the control of a higher-level communications processor or as a peer of other intelligent switches. Remote
concentrators control a community of terminals or distributed application processors. gathering, queuing, and
multiplexing their transmissions onto one or more high-speed network trunks. These concentrators also often
provide protocol conversion and gateway functions to attached devices.
Point-to-point multiplexing of a carrier circuit is useful in sharing costs where there is a co-located community
of users who need access to a common resource elsewhere. Unfortunately, this simple situation is not pervasive
in business; complex patterns of access are now the rule. Furthermore, higher-capacity circuits often need a
number of users to be justified.
Figure “Transit Routing and Switding Techniqzres” shows that it is possible to concentrate traffic onto a trunk line
by bringing some traffic to one trunk termination from locations even more remote (locations “C” and “D” on
the figure). This would also provide users in that remote location with access to ant’ information resource
located at the near trunk termination (“A” on the figure). But to support this config&ation, there must be a
network device at location “A” that can distinguish between traffic from “C” or “D” which is destined for “A”
and then destined for location “B.” This requires a routing or switching function.
Switching devices are the core of modern data networks, because the collateral need s of concentration of
traffic for ecoi jomy aiId ful I connect ivity to support all types of information connectio ns cannot be readily met
without them. Private data networks can be defined as networks in which user devices provide this switching,
and public networks are those where the switching is provided by the carrier as a part of the service.
Switching devices are often called “nodes” because they form the junctions between routes or trunks in a data
network. Because the control of node operation and routing of information will control overall information flow,
the “management” of a network is tvpically
r’ based on the management of the nodes within it,
communication at all, resulting in nothing to monitor. When this happens, or where a component of the network
is to be examined outside a connection dialog, an active testing device is needed. -
The simplest testing devices generate patterns of data. either to display on a terminal or to “loop back” to the
source for comparison. These are often called “bit error rate” or “bit tine error rate” testers. Most are similar to
interface monitors in size and are restricted to use with very simple protocols, since they tack the intelligence to
obey complex rules for information exchange.
Data line monitors, with microprocessor intelligence, provide active testing capabilities of even complex
protocols. Called “protocol emulation,” this testing capability allows a user to certify the basic operation of a
device and confirm the essential characteristics of a carrier service.
Testing devices can sometimes be used by inexperienced personnel, providing that the device has a simple
“go/no go” indicator. In most cases. however, testing devices have the same constraints on usage as apply to
network monitors.
Data communications and data networks are more than products and services. Personnel must plan. select,
install, and tnaintain equipment and coordinate services. Because data networks project computer information,
they are closely tied to computer planning and operations. Because they may utilize carrier facilities like Tl in
conjunction with voice services, data networks are also often linked to telecommunications planning and
operations.
Many businesses have computer and communications organizations that have been inherited from a period
when the role of data communications was very different than it is today. In some cases, this results in a
subordination of data communications issues to data processing interests. In others, a “runaway” network plan
has little relationship to the computers and terminals that must be served. But in most cases. defects in
organizational coordination hamper the diagnosis of problems and the support of users.
As data networks become more critical to business, they must be placed in a reasonable planning and support
context. Computer technology and wide area transport services are becoming less expensive daily. If the
enterprise of the future is to be dependent on networks, cost-benefit constraints of networks must be addressed in
planning. and the needs of the networked enterprise must be met in technical support.
Planning Considerations
Companies traditionally plan network services based OII a set of demands presented by the information
technology and computing planning tasks and the constraints set by the carrier service structure in place in the
areas to be served. This presumes that the network is relatively flexible to meet any information technology (IT)
goals and that computer technologv d is relatively inflexible. Cost trends, cited above, clearly show this to be
false, even in the present.
In modern IT planning, network constraints on the computing systems relationships are considered as early in
the process as computer technology and sofiware constraints. The goal of most businesses is the enterprise
network, a collection of communications services that will meet the information transfer demands for all types
of information, now and in the future.
Because computing relationships are changing rapidly in the face of plummeting desktop technologv costs,
networks are increasingly supporting client/server and peer communications models. As shown earlie;, this
creates a need for increased connectivity along the “fringes” of the network. The enterprise network of the future
is therefore likely to have fewer preferred information paths and be much more dependent on nodes and shared
trunks than was the case only a decade ago.
Increasingly, users require higher-capacity network services that are priced according to actual usage
(connection time and/or bandwidth allocated) rather than on the peak capacity available. New technologies, such
as switched digital (n x 64K bps) transmission, frame-relay data transport, Switched Multi-megabit Data Service
(SMDS), and ATM, are enabling such services to evolve.
Broadband networks of the future will probably have a stronger carrier component, much like virtual networks
of today, because of the need to concentrate the traffic of many users to create econom ical service connections.
The carrier component of broadband services is evolving from Sonet and BISDN principles. Future broadband
networks will also require local distribution of high bandwidth traffic. The local distribution strategies are
evolving from hub and FDDI principles. Somewhere, the two trends must mesh.
As users plan for the future, the role of carrier “virtual” networks for data communications will also become
more important. A virtual network offers users the services of a private network without the equipment
investment. This insulates the user against the problems of change, but it places the burden on the carrier. Users
who are totally dependent on virtual services may find that those services cannot be adapted rapidly to take
advantage of new opportunities. A balance of risk and opportunity will be required to support the best
competitive business communications and information processing structure.
The problems of change are magnified by the fact that new information technology plans often disperse
computing and information storage resources throughout the enterprise, and the user may no longer have stable
and comfortable user-to-host relationships. When a problem occurs on an application that “slaves” a user to a
particular computer, computer operations often receives all the support calls on that application. Where the
network insulates the user from the information source, perhaps even to the point of m ak ing host access
transparent to the user through distributed or client/server relationships. the “network” iS now the focus of
problem reports.
The concept of a central support facility or “help desk” is not new, but it is of increased importance. Without
such a support resource, end users simply cannot locate the proper point of contact to report problems without
actually entering into problem isolation procedures themselves, a move that is impractical in the light of the
increased technical complexity of IT systems. But help desks also provide a value to the technology
infrastructure of a company. by coordinating technical personnel during problem diagnosis and isolation and by
providing a central point for vendor contact and follow-up.
The organizational impacts of a central support desk can be considerable, but so can the technological
implications. Integration of network management systems into a single platform is valuable where data networks
are centrally supported, but it is critical if data network management is to be viewed as an element of user
information services management. This view is essential to any central support of IT resources. In the longer
term. it may be very desirable to integrate the management of networks with the management of the computers
and terminals they serve. The user is then relieved of the need to isolate application problems from delivery
problems. and the overall effectiveness of the support organization in maintaining application availabilit\r r’
improves.
This report was prepared exclusively for Datapro by Thomas Nolle, President, CIMI Corp. CIMI Corp. is a technology firm located in
Voorhees, New Jersey, and specializing in strategic planning and market development. Tom’s views on new communications products and
services are regularly quoted in major trade publications.
CIMI Corp. provides strategic planning, market research, market forecasting, competitive analysis, and systems integration services to
users, vendors, and communications carriers.
Figure 1. Time
Analog and Digital Coding Tecttniques
Synchronous Transmission
/I Application
1
RWentatiOIl
Presentation I
Session session
hpofi
Network
Network
Network
44 Network
Host/Terminal
Peer Relutlonship
Relationship
Client/Server Relationship
Intern&work Relationship
Figure 5.
Transit Routing and Switching Location ‘C’
l
Techniques
\
8
D
Y
\\ \\ /
\ \ \L 8
A
Location ‘B’
A “Switching or Routing’Node
lb
Figrvc 6,
Network Monitoring ami
Management
nt