Network Basics

Woodard • Gattuccio • Brain

Networking Basics

Networking Basics by Shay Woodard, Nick Gattuccio, & Marshall Brain

The following article is excerpted from Chapter Two, “Networking Basics,” of the book Microsoft Technology: Networking, Concepts, Tools, by Woordard, Gattuccio and Brain (Copyright © 1999 by Prentice Hall PTR, Prentice-Hall, Inc.) Reprinted by permission.

A network is two things – technology and business. While the first of these is quite obvious, network technology is sometimes so overwhelming to non-technical managers that they lose sight of its basic role – to perform as a business tool. This is a classic case of losing the forest for the trees. While network technology is very important, its complexity should not obscure its role as a business tool. In this chapter, our goal is twofold. First, we hope to demystify network technology by painting a detailed picture of its key pieces and defining the technology’s key vocabulary. While a great deal of technology is involved, it all comes together in a straightforward way. Understanding the vocabulary clarifies nine-tenths of the technology. In Chapter One we introduced network technology in terms of the network services it provides. While this is the plan we’ll follow throughout the book, it is nevertheless important to understand network technology nuts and bolts. Our goal in this chapter, then, is to address network technology at its most fundamental layer – the wires, data, and connections. This chapter’s second goal is to highlight the central business issues that fall under the rubric of managing technology. In few other areas of business planning is there greater risk of technical systems turning the tables on managers and creating a situation where the technology dominates business. The two – technology and business – work hand-in-glove because a poorly implemented network creates limitations and constraints that limit business planning options.

PART ONE: The ABCs of Network Technology The basic unit of network organization is the LAN – the Local Area Network. This is very important. Nearly everything in network architecture and design builds on the fundamental principle that a LAN is the network building block. While in practice, LANs are typically broken up into smaller units – LAN segments or workgroups, for example – this does not alter the LAN’s status as the fundamental unit of network order. It is also helpful to think of LANs not as computing technology, but as communication technology. Their central function is letting users share computing resources by allowing data to flow easily from device to device. On this simple principle, designers use cable to connect computers and other network devices (printers, for example) within a limited area (for example, within a department or on a floor of a multistory building). Then, using these LANs for building blocks, networks can expand to virtually any scale by simply linking together the individual LANs. This allows you to link floors of a building (each with its own LAN) into a company-wide network, or link several buildings on a campus, or even link up sites around the world into a powerful enterprise-wide network. This LAN-based model lets you build networks on virtually any geographical scale.

Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 1

Woodard • Gattuccio • Brain

Networking Basics

Communication Media & the Network Interface In a network, information (data) traverses a communications medium (typically a wire) as electrical signals that originate and terminate in computing devices like computers and printers. To visualize the process, picture a busy freeway as seen from the air. Traffic flows smoothly and vehicles enter and exit at ramps located every few miles along the route. Viewed from above, the traffic appears to be a single continuous stream. Upon closer inspection, however, you notice that vehicles enter and leave the freeway in a seemingly random fashion. But this is not the case. The motion is not random. Each vehicle has its own point of origin and its own destination, irrespective of the streaming flow of traffic. The analogy is not perfect, but it successfully isolates three basic elements of a network: its communication medium (typically a cable, but radio frequency and infrared transmissions can also serve as a data communications medium); its data (which flows in its most elementary binary form – essentially zeros and ones); and the points of contact between network wires and the computing devices to which they are connected. These latter are called network interface cards (NICs). These three are the highway, the traffic, and the entry/exit ramps. (There is a fourth critical piece on this layer, the connection devices like hubs and switches; we’ll discuss these shortly.) Cable is the communication medium over which LANs pass data – the electronic highway. As with the transportation analogy, the medium can take many forms (freeway, boulevard, street, or road). Network cables are typically coaxial, twisted-pair, or fiber optic. Some are shielded – that is, insulated to help prevent electromagnetic interference from corrupting the signal – while others are not. In twisted-pair cable, two insulated wires are twisted, then encased in a sheath. Coaxial, on the other hand, has a solid wire at its center which is surrounded by insulation, then an outer metal screen. Both transmit data in the form of electrical signals. With fiber-optic cable, however, data is transmitted as modulated light waves traveling through a glass medium. These different types of cable vary in cost and transmission capacity, and each typically serves a specific network role. For example, on LANs, unshielded twisted pair (UTP) cable is most commonly used because it is inexpensive, easy to work with (it is thin and flexible, much like telephone wire), and it performs well over a typical LAN’s short distances. On the other hand, a high-speed backbone (high-capacity cables that connect LANs together) might use a fiber optic cable because of its large transmission capacity. Fiber optic cable has the added advantage of being non-conductive, which is advantageous for external cables (between buildings, for example) that are vulnerable to lightening strikes. Coaxial cable (thick, stiff cable similar to that used for TV-to-VCR connections) is rapidly losing favor on modern networks because it is difficult to work with and integrates poorly with switched networks. Network devices make their physical connections to network cable through a special adapter called a network interface card (NIC). The network transfers data over cable in its most elementary form – as a stream of bits (zeros and ones). The NIC takes in data and converts it to a form that fis the transmission medium – the cable. Computers hand their data over to the NIC, which chops them into small segments and puts them into “envelopes” called packets. Then, the NIC moves the packets onto the network. The NIC also does the reverse – picks incoming packets off the network and hands them over to the computer. NICs are standard on many computers, printers, and other devices designed for network use, or else one can be installed. Data passes over the network inside electronic envelopes called packets. (We’ll discuss packets and packet switching in more detail in Chapter Ten.) Quite simply, though, when a user at one workstation wishes to exchange information with a user at another, network software on the first user’s workstation places the data into one or more packets. Packets have a limited size, so if the “message” is too large to fit in a packet, the software breaks it up into manageable fragments and places each into its own packet. The packet also contains special information, including its destination address. Then, the NIC at the sender’s Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 2

Woodard • Gattuccio • Brain

Networking Basics

workstation dispatches the packet onto the LAN. As the packet transits the network, every device attached to the local network “sees” it. If the device is not the intended destination, it ignores the packet. If, however, the address on the packet matches that of the network device, the NIC at the receiver’s end copies the packet and removes the data from the packet and hands it over to the computer, whose own software takes over. The above illustration simplifies the exchange somewhat, but it nevertheless paints a sensible picture of how a network functions at its basic level – at the layer of the wire. The principles of packets and addressing, and the view of a network as fundamentally wire, and the role that software plays in adding “intelligence” to the flow of network communication are among networking’s core principles.

Moving Data: Ethernet & Fast Ethernet, etc. Data needs an engine to drive it. While the wire itself is the transmission medium, different technologies handle the task of physically moving the data over the wires and regulating the characteristics of the electrical signals that carry it. The most common of these is Ethernet, in place in a reported 80% of LANs. Other technologies include Token Ring, FDDI (fiber distributed data interface), LocalTalk (for Apple Macintoshes), and some others. Each of these standards governs several key features of how data moves around a network. Two technical aspects involve the physical characteristics of the electrical signals themselves, and the manner in which the data is packaged (the look and feel of the packets we spoke of). Other issues involve the speed at which data can flow, the physical length of wire it can support, and the network’s topology (that is, the physical layout of the network, as in a star, ring, or straight-line bus). Finally, there is the very important issue of how it brokers shared access to the network medium – the cable – to avoid data “collisions.” Ethernet is the most popular option for controlling this physical layer of data flow because it offers an excellent balance of cost, speed, and ease of implementation. Relatively inexpensive, Ethernet supports data flowing at a rate of 10 million bits per second (10 Mbps), although an improved Ethernet, called Fast Ethernet, supports speeds up to 100 Mbps. Its main weakness shows up on busy or crowded LAN segments, where data “collisions” may degrade network performance. This may occur because the access method Ethernet uses to push data onto the network allows more than one station to access the wire at the same time. If this happens, a collision will occur. Ethernet corrects the collision by prompting those involved to wait briefly, then re-send their packets, but the error and recovery require time. On a very busy network, frequent data collisions may degrade network performance. Ethernet uses an access method wherein each device, or network node, initiates its own access to the network. When a workstation, for example, wants to send data over the network, its NIC “looks” to see if the line is free. If it detects another signal on the wire, it waits a short interval, then tries again. When it sees the line is clear, it sends its packets. The problem occurs when more than one network node tries to access the network at precisely the same time. Both see a free network, so both dispatch their packets, which then collide. An alternative to Ethernet is a standard called Token Ring. A Token Ring network avoids collisions altogether because it uses a different access method that doesn’t allow devices to initiate their access independently. In a Token Ring network, a special data packet called a token passes around the ring from device to device in just one direction. If a device wishes to access the network to transmit data, it must first wait until it receives the token. It cannot initiate access without it. Once it obtains the token, it dispatches its data, along with the token, out onto the ring. The destination receives the message, pulls it from the network, then releases the token and the process repeats. On a corporate LAN, Fast Ethernet is typically the standard of choice – in part because it is the market leader, and in part due to its great value as a cost-effective, relatively high-speed option. Beyond the scope of LANs, however, networks may require even higher-speed connections and more reliable data flow. These Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 3

Woodard • Gattuccio • Brain

Networking Basics

include network backbone connections that link geographically separated LANs. (Short-run backbones – for example, those connecting LANs on separate floors of multistory buildings – can function well with Fast Ethernet.) Linking multiple buildings, as with several buildings on a university or corporate campus, are typically served by a robust connection like FDDI. Standing for Fiber Distributed Data Interface, FDDI offers 100 Mbps connection speeds that are highly stable (i.e., resistant to electromagnetic interference or other line “noise”). FDDI is also appropriate for peer-to-peer server connections, particularly where a large, busy network is supported by multiple groups of pooled servers. It should be noted that in actual practice none of these technologies delivers their rated speeds. Information in the data’s envelope (the packet headers and trailers) increases by about 30% the data’s “packet overhead.” Add to this gaps between packets and data collisions (with Ethernet) and a 10 Mbps Ethernet network is able to sustain throughput of only about 2.5 Mbps. So far we have presented a very simple picture of a network that consists of devices, data, NICs, and cable. This picture is based on a simple model, called a bus network. In such a LAN, the cable is viewed as a single line with devices connected at various points along the way. Such a LAN bus topology is rarely used any more, but its value comes in how well it illustrates the interaction of the basic network elements. While a bus topology itself is archaic, its pieces and the ways in which they interact are not.

Figure 0–1 Basic LAN Bus Network Basic LAN Bus Network Topology. This simple topology illustrates how the fundamental “wire-layer” network pieces – device, data, NIC, and transmission medium interact to complete network connections. While the bus topology is outdated, its model of network interactions is not.

The concepts exposed in Figure 2-1 apply to more common topologies as well. While significant differences separate the topologies of LAN bus, Ethernet, and Token Ring networks, the manner in which their basic pieces interact remains largely consistent. Network professionals may bristle at this last statement, as technical subtleties tempt one to cite exceptions and qualifications to such a blanket statement. For the purposes of this overview, however, the picture is reliable – but with one notable exception. So far, we have not mentioned one of the most important elements of a LAN – the hub. The hub, along with other network connecting devices, enables networks to expand. If LANs are a network’s basic building blocks, then the central connectors are the glue. But this is intelligent glue.

Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 4

Woodard • Gattuccio • Brain

Networking Basics

The Central Connections: Hubs, Bridges, Routers & Switches We said earlier that LANs are the basic building blocks of a computer network. Large-scale networks knit LANs together with cables (or other transmission medium). But linking LANs together introduces complexity, particularly if the network has a great many LANs, or if one or more are separated geographically. As we saw in the bus model, all of a network’s devices share the communications medium (the cable) in such a way that every “message” passes every device, and every device “looks” at every message. On a LAN’s small scale, distributing all of the traffic throughout the network does not typically overwhelm it. When several LANs are joined into a large, wide-ranging network, however, the potential for problems is enormous. It would be extremely inefficient for a 50-LAN network, for example, to distribute all of its traffic over every one of its segments. A much better system would build on the data packet’s inherent addressability. Recall that data packets contain the data’s destination “address.” But while packets contain the address of their destination, the packets themselves have no way of using the address information. That’s the NIC’s job. Instead, packets simply traverse the network. The NIC pays attention to passing traffic, and when it notices that a packet is addressed to its attached device, it copies the packet off the network. The packets themselves have no “intelligence.” Another problem is that LANs have physical limitations. There are limits to how many workstations a LAN can efficiently support, and to how much traffic it can carry. There is also a limit on how far a run of LAN cable can extend (about 100 meters) before its signal degrades. The point is, building large-scale networks is constrained by technology that only performs well on a small scale – on the scale of the LAN. These issues create an enormous technical constraint on network scalability. Without solutions, large-scale networks would be impossible to build. Fortunately, these problems are solved by the technology of central connection points – hubs, bridges, routers, and switches. These devices allow networks to be segmented, and they also add “address-intelligence” at the level of the physical wire. This makes the potentially overwhelming traffic problem manageable. They also solve the problem of line-length constraints – where the physics of data transmission over wire degrades the electrical signal over distance – by acting as repeaters. Hubs A hub is among the most critical elements of a LAN, as it is the point of central connection for all of the LAN’s shared devices. A typical hub has multiple ports to which a LAN’s devices connect. The actual number of ports and the transmission capacity of each varies widely, as does their price and additional features. Regardless, a hub serves the same function as the shared cable in the bus model – it connects devices on the LAN. Figure 2-2 depicts the hubbed LAN.

Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 5

Woodard • Gattuccio • Brain

Networking Basics

Figure 0–2

A Simple Hubbed LAN

Rather than a straight-line connection, as with a bus topology, a hubbed LAN relies on a hub to function as a central connector. Traffic from any device on the LAN goes to the hub, which “repeats” the message out to all of its connected devices.

On a hubbed LAN, devices exchange data by first sending it (through its NIC) to the hub. The hub, in turn, repeats the “message” back out to all of its connected devices. Again, as with the bus LAN, only the device to which the message is addressed will copy the message, while the others ignore it. By repeating the message, the hub extends the effective distance the message signal can extend, since the hub transmits what is essentially a fresh, new signal. This way, every device on the LAN can be as far away from the hub as the connecting medium will allow (100 meters). That probably wouldn’t be practical, but it might. What’s important is that this lets a LAN scale up very easily, since adding new devices does not increase the physical length of the LAN (as adding devices to a bus would), but only adds a new connection to it. If a LAN grows to the point where all of its hub ports are in use, then you can simply add a second hub to one of the original hub’s ports. The second hub repeats the first hub’s transmissions through all of its ports. Hub jumping like this allows enormous flexibility and scalability in LAN design, although there are physical limits to how many jumps are allowed. The effective scale of a hubbed LAN is also constrained by network traffic. Because hubs repeat all transmission to all of its connected devices, hubs can easily support a volume of network traffic that exceeds the cable’s ability to carry it. Exceeding the medium’s effective transmission capacity will degrade network performance. When a LAN reaches its maximum effective size, it is necessary to install another LAN and then connect the two. From another perspective, installing another LAN allows designers to segment a network, effectively breaking it into smaller pieces, improving performance while adding even more scalability. Hubs connect devices within a LAN. Three main types of devices are designed to connect internetworked LANs – routers, bridges, and switches.

Routers A large network that links, for example, 25 LANs, might support five hundred or more users. Under the hub model, where data is broadcast to all connected devices, a network such as this would collapse under the weight of its traffic. To scale beyond the LAN, there clearly needs to be a way to segment and route network Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 6

Woodard • Gattuccio • Brain

Networking Basics

traffic. A large network needs to have intelligence on the wire so it can read data-packet address information and use it to segment and route its network traffic.

Figure 0–3

Inter-Networked LANs

Using “intelligent” connectors like routers and switches, virtually unlimited numbers of LANs may be inter-networked into enterprise-wide networks. Such networks are not limited to just a single geographical location, but can unite a network even on a global scale.

It is reported that typically 80% of transmissions on a workgroup or departmental LAN are destined for stations within that LAN. Filtering out that 80% and keeping it off the larger network lets designers build both scale and efficiency into a network. Routers are among the most popular technologies to filter network traffic. A router is a device that, like a hub, has ports through which data passes. With a router, however, data passes only from one LAN to another. Unlike hubs, routers do not blindly repeat the data they receive. Instead, routers are “intelligent” devices. As data packets flow into them, they inspect the packet’s header (which contains the packet’s destination address) and based on this information they make a routing decision. The most obvious decision the router makes is when the destination address is on the same LAN as point of origin. In this case, the router stops it from leaving the LAN. But a router is capable of making more than simple yes/no forwarding decisions. Because routers understand communications protocols (like IP, which governs Internet addressing), routers can make configurable routing decisions. For example, administrators can use routers to help balance network traffic by configuring them to use predetermined pathways. Because routers are configurable, they lend themselves to complex uses, as in a network security system. Routers can play an important role in a network’s firewall, for example. Another strength is that routers can connect LANs that use different protocols – for example, connecting an Ethernet LAN with a Token Ring LAN. The tradeoff for all of this intelligence and functionality is twofold. First, routers are typically more expensive than other inter-network connection devices, and they are also more complicated to set up.

Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 7

Woodard • Gattuccio • Brain

Networking Basics

Bridges Bridges are simpler and less expensive than routers, but offer similar inter-network capability. Typically, bridges have ports connected to two or more separate LANs. Rather than complicated routing decisions, however, bridges make straightforward yes/no decisions about forwarding the data packets they receive. Bridges base their decisions on the packet’s destination address, which it compares to a stored table of known network addresses. If the destination is located on the same LAN from which the packet originated, it will not be allowed across the bridge. If addressed to a device on another LAN, the packet is bridged to a port through which it can reach its destination. Like routers, bridges inspect information in the packet’s header. But because they make more limited use of packet information, they look at far less of the header information than routers do. This makes them very fast. They are also protocol-independent, not because they “understand” protocols, as routers do, but because they can ignore protocol issues altogether. This makes them extremely versatile on a homogeneous network. While not as sophisticated as routers, bridges do possess some filtering capabilities. Switches Switches are the most recent, most sophisticated and most expensive of inter-networking devices. Like routers and bridges, switches also link LANs. Their strength comes in their ability to link many LANs. They are like bridges, but because they have multiple ports they can manage complex switching among multiple LANs. This makes them useful tools for segmenting network traffic. Switches can also allocate bandwidth in their segment connections. This allows network administrators to balance network traffic to reduce congestion on workgroups while still providing high-bandwidth “pipelines” where they are needed (in connections to server pools, for example). Switches provide the same kind of address-intelligence (filtering and forwarding) that routers provide. But rather than working with the IP address, as routers do, a switch uses a special media access control (MAC) address that works at a very low level, where the network devices interface with the medium (cable). Network nodes, or devices, are mapped to unique MAC addresses, and the switch can then handle routing and filtering tasks independent of protocols or other constraints. Among the greatest strengths of a switch is its ability to support connections across multiple LANs as well as supporting simultaneous transmissions. For example, a switch might have a high-bandwidth connection (100 Mbps) coming in from one LAN which it switches to ten different 10-Mbps segments of a second LAN. A switch is capable of servicing traffic to the ten smaller connections simultaneously. This offers enormous efficiency and speed. Speed can be even further increased with a technique called duplexing, which can effectively double the bandwidth of a connection. Our summary of these central connectors is necessarily brief. The technology is in many cases very complex. It is also rapidly evolving. There are variations (like self-learning hubs) and powerful hybrid technologies (switching hub, for example), but the concepts remain relatively straightforward. Summary of Internetworking Networking technology has evolved rapidly in recent years, and the trend is almost certain to accelerate in coming years. In its early days (the early 1980s), networks were essentially local – they were LANs. As the role of the personal computer in business (and the university) mushroomed, LANs proliferated. The power and utility they offered by enabling a shared computing environment was so compelling that the next step in their evolution was inescapable – inter-networked LANs.

Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 8

Woodard • Gattuccio • Brain

Networking Basics

This class of network technology, the central connectors, are the essential feature in the evolution of networks. They are the cornerstone of network scalability. As their technology evolves (expanding the power and functionality of the central connections, and increasing their speed and bandwidth) the shape and characteristics of networks themselves will also continue to evolve. Computing Devices On top of the network layer that we’ve described so far sit the computing devices themselves. These include workstation computers of many types (including portable, or “laptop” computers), server computers, and a wide assortment of peripheral devices like printers, data storage devices, and many others. In the course of discussing network services throughout the remainder of this book, we’ll discuss computing devices extensively. Although the subject of computing devices is vast and far-reaching, from the point of view of the network, computing devices are very straightforward. They are simply points of origin and destination for network data. When a workstation sends a print job to a workgroup printer, for example, all the network layer sees is the stream of data. It doesn’t know the data is a print job, and it doesn’t care. Database file, print job, email message – these are all the same to the network layer. This is an important feature of networking – its layers – and we’ll come back to it at several points throughout the book. Erecting network technology (and computing technology in general) in discrete layers like this allows layers to evolve independently of one another. For example, while to a certain extent applications need to be “network-aware,” in general a word processor doesn’t care if it is on a network or not. These layers – network, computing devices, and software – enjoy a reasonable degree of independence, as does their implementation in a network environment.

Bringing It All Together In Chapter One we discussed networks in terms of their architecture – client/server, peer-to-peer, etc. We outlined network functionality in terms of the services a network provides. The overview in Chapter One, with its emphasis on architecture and services, lays the foundation for how we intend to proceed throughout this book. At the same time, the technical viewpoint (at the level of the nuts and bolts, if you will) is an equally legitimate window on network technology. While we’re electing not to pursue it at length beyond our brief treatment here, it is nevertheless central. These two windows into networking (the bottom-up and the top-down views, in a sense) interact very closely. In fact, they are inseparable. A network’s architecture, and the extent of the services it provides, depends on the state of networking technology. Where the network imposes constraints (as with bandwidth), the constraint extends to every piece on the network. This point is central. Our transportation analogy carries over to this point. Transportation technology (vehicles – their design and functionality) reflects the state of transportation infrastructure. If trucks, for example, were twenty feet wide and a hundred feet long, apples would be cheaper. While it would be easy to build trucks like this, it doesn’t happen because the network medium (roads and freeways) can’t handle trucks like this. The network’s constraints become the technology’s. And the point works in reverse. Supersonic flight, for example, is extremely expensive and environmentally unsound (due to noise issues primarily). These are technological constraints. Consequently, our commercial air-traffic network reflects these limits of aircraft technology. Networking technology is evolving at a more rapid pace than any other aspect of business systems. Advances over just the past decade have created communications and commercial platforms that only a few could have imagined in the 1980s. There is no reason to doubt the trend will continue. The software, hardware, and communications companies leading the technological vanguard are largely prosperous, and venture capital is abundant. The climate for technological innovation is optimistic, even frenetic. On the Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 9

Woodard • Gattuccio • Brain

Networking Basics

marketing end, the industry is consolidating its resources, as software, communications, media, and entertainment corporations merge into multi-media giants, poised to exploit new opportunities in the field – opportunities that will be fostered largely by advances in technology. Enhancements to networking infrastructure are occurring at an extraordinary pace and on a global scale. Traditionally separate mediums – computing, broadcast television, telephone systems, cable systems, and even satellite technology – are rapidly crossing one another’s boundaries and integrating their respective technologies. The common denominator in all of this is networking.

PART TWO: Managing Technology The common complaint is that the technology manages the business. This exaggerates the point, of course, but the point merits attention nevertheless. Poorly managed, technology can hold a business hostage. The day is long past where business managers could foist technology decision-making onto technical management and concentrate instead on business. Among our central themes in this book is that technology is business. What things does a business manager need to know about technology? Does he or she need to know the difference between Cat 5 and Cat 3 cable, or the relative merits of a bastion host versus packet-filtered firewall? Probably not. But a business manager does need to know the right questions to ask. When faced with any technology decision, here are some of the fundamental business questions a manager should be asking: •

What does this technology do for the business? What does it not do? How will our business systems benefit, and how will they be constrained?

•

How much does this technology cost? What is the learning curve for users? What are its risks? How does it compare with alternative technologies? Where does the new technology stand compared to its predecessor – is it an incremental improvement or a significant innovation? How will its introduction impact other pieces of the network?

•

How will the technology affect our business partners? Our clients and customers? Will it increase our ability to communicate with them? Will it enhance the business systems that support our interactions with them, or will it degrade them, or have no effect?

•

If this technology breaks, what happens to the rest of the network, and to the business systems it supports? How easily and inexpensively can it be repaired? Can the technology be made fault-tolerant, and at what cost?

•

How difficult is it to administer and maintain? Can existing technical staff manage it, or will they require training? If so, how much? How great a leap in technical expertise will it require?

•

Is the technology mainstream, or is it extreme? Will it be widespread in two years? Will it still be supported then?

•

What are the recurring costs of the technology? What are the exposures? What is the useful life of the technology?

•

How does it compare with network strategies employed by other companies that are our size? What about other companies in our industry?

•

How do you calculate the return on this investment? What is the return? Will benefits accrue to the company immediately, or is this part of a long-term strategy that is tied to the network’s (or the business system’s) long-term evolution?

Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 10

Woodard • Gattuccio • Brain

Networking Basics

There is no perfect technology, nor perfect mix of technologies. Like building anything else – a house, for example – any number of sizes, shapes, and designs can suit a business’s needs. It is essential to understand a technology’s goals and purposes before understanding its technology.

Network Cost Centers In Chapter Three we discuss network management and administration by dividing networks into three service zones – the network, the data center, and the desktop. (See Chapter Three for extended definitions of the three service zones.) The goal there is to simplify a network’s vast array of technology by logically dividing and classifying its parts in manageable segments. Management’s view is equally well served by such a compartmentalized approach. The key issues in managing the network (as opposed to administrating it) center on asset management, cost management, and best practices. We cover administration best practices and asset management in Chapter Three, but we’ll discuss higher-level business management practices, as well as cost management, briefly here. What are the key cost centers for a network? In fact, the list is long. Many of these are obvious (direct budgeted costs), while many others are hidden or indirect costs that may not find a line on the IS budget. Some are recurring (like maintenance costs and communication fees) while others are cyclic (like training for rollout of new technology). Below we provide an outline of a network’s primary cost centers. Each should be assessed for each of the network’s three zones. Hardware The category includes expenditures for new (and leased) equipment. In the data center, this includes server computers, data storage equipment, central communication equipment and related peripherals, as well as environment control devices (temperature and humidity) and power conditioning and management devices. For the network zone, this includes core infrastructure items like cabling and switching equipment (hubs, routers, etc.). The desktop is the most dynamic area for hardware costs in most companies, including desktop and mobile computers, internal devices (hard disks, other read/write components, optional circuitry like video cards), and peripherals like printers and scanners. The category includes secondary acquisition costs (e.g., lease fees) as well as disposal costs. Some costs, like memory, storage capacity, etc., might apply to hardware in more than one zone. Software Server software, operating system software, management tools and utilities, security software, development software (e.g., programming languages, development environments, etc.), and general utilities form the bulk of costs at the data center. At the network zone, typically software costs are minimal, although switching and routing software may fit here. The desktop is, again, where costs can be most fluid and unpredictable, and most difficult to manage. Included here is new software, upgrades to existing software and annual site license fees. Applications themselves that populate desktops range from clients applications working with core server systems to business applications and utilities. Consumables Network-related supplies, including storage media (disks and tapes), printer supplies, etc. This category typically does not apply at the network zone.

Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 11

Woodard • Gattuccio • Brain

Networking Basics

Network Management & System Administration This is a very large category which may span two or more zones depending on the network’s configuration. Following is a summary. •

Asset inventory management. Costs for maintaining asset inventories, periodic inventory taking, change management, and maintenance of inventory tools, like databases and other records. Costs are primarily labor.

•

Application management. Includes server and desktop applications, custom applications, operating system, utilities, and other application services. In the case of custom applications, this includes configuration and maintenance, not development costs. The category includes labor, primarily, for configuration and access control. Includes managing upgrades and deployment of new applications. Costs here can be significant. For example, an operating system upgrade generally requires that each machine on the network receive on the order of 4 hours of personal attention.

•

Hardware configuration. Costs to deploy and configure hardware on the network, change configurations, upgrade existing equipment, modify network topology, etc., as well as disposal of retired equipment.

•

Storage management. Includes costs for capacity planning, developing backup and archival systems, disaster planning, and data-access systems (for data recovery). Includes planning integration of hardware devices, as well as disk and file management, routine storage-media maintenance (defragmentation, testing, etc.), record keeping, and policy planning.

•

System planning. Costs for ongoing monitoring and managing of network systems, periodically reassessing performance, tracking growth, and planning evolution and expansion of network capacity and performance.

•

Procurement. Costs for evaluating and testing technology (both hardware and software), developing options and contingency plans, and documenting purchase recommendations. Includes the issue of lifecycle management.

•

Security management. Developing, implementing, and monitoring network security systems, policy, and procedure. Includes password protection and access control, network gateway security (i.e., firewall), and virus protection.

•

Traffic management. Costs for monitoring and assessing network traffic, planning load balancing, periodically reassessing traffic patterns, and optimizing network traffic flow.

•

Performance management. Closely related to traffic management, but expands to include non-traffic-related performance issues like software, operating system, and application services, as well as network hardware, and fine-tuning configurations to optimize network performance.

•

Troubleshooting & Repair. Unscheduled maintenance on network and data center hardware and software to correct malfunctions. (Troubleshooting and repair at the desktop zone is covered in the Support Costs section, below.)

•

Backup (spare) equipment and unscheduled replacements. Critical network components, whose failure may negatively impact the network, may have replacement units on hand to allow rapid replacement in emergencies. Furthermore, equipment failures that can not be cost-effectively repaired may require replacement. Warranties or service contracts may negate the latter.

•

Recurring costs. Costs for service contracts and other incidental recurring expenses.

•

Consulting & Outsource expenses. Costs for outside services contracted to augment in-house administration and management.

Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 12

Woodard • Gattuccio • Brain

Networking Basics

Communication The cost of connectivity, which is typically applied at the network zone. These costs include fees for things like ISP hosting, WAN connections, leased-lines, on-line access, remote access costs (long-distance tolls to RAS server), etc. Support &Training Training and support costs, more than any other, typically carry hidden or incidental costs. Furthermore, a great many unknowns (for example, predicting needed support for an operating system upgrade) typically fall into this category of cost. For these reasons, budgeting, managing, and controlling costs in the area of support and training are among the most challenging for business mangers. Following are some key types of support and training costs. Note that training costs can involve end users, managers, and/or IS technicians. •

Administration and management of support and training services. These include labor costs for both management and administrative support.

•

Vendor management. Costs associated with contracting for outside training and/or support services.

•

Course development. Costs (labor and materials) to develop training course curricula and materials.

•

Training facilities, materials, and logistics. Costs for maintaining (or renting) training facilities, printing or otherwise generating course materials, and logistics for providing equipment (e.g., computers), transportation, or other needs for both trainers and trainees.

•

Outside training. Costs for sending technicians, managers, and/or employees to a remote training site – includes course costs, lost productivity, transportation, food and lodging, etc.

•

Helpdesk support. Labor, training, and administration of help desk systems.

•

Troubleshooting & Repair. User problems, either hardware or software related, which cannot be solved through help desk intervention typically require a service call.

Development Costs Includes the cost to design and develop, test, document, rollout, troubleshoot, and fine-tune custombuilt applications. In cases where application development is handled by a third-party, this includes all associated costs. Downtime Costs Frequently omitted from technology budgets is downtime costs – both scheduled (e.g., for maintenance or for the rollout of new technology) and unscheduled (due to equipment failure). Summary The above sections summarize typical categories of costs a company typically incurs to manage, support, and maintain a network. Most lend themselves to subdividing along the network’s service zones – data center, network, and desktop – which helps isolate costs at a more granular level. But there are two levels to profiling costs for a business network. Only the first involves identifying network costs, as we’ve done above. The second is managing costs.

Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 13

Woodard • Gattuccio • Brain

Networking Basics

Managing Network Costs The total cost of ownership (TCO) is a concept frequently used to understand the cost of computing services. Simply put, TCO is a method that aggregates the lifetime cost of computing equipment (including such things as hardware, software, installation, service, support, training, upgrades, downtime, and anything else that may apply under special circumstances) as a cost-analysis metric. The method gained currency in the hands of manufacturers of “thin client” systems to help them make the case that such systems were superior business investments than “fat” clients – typical workstation computers. (Thin-client systems hark back to the mainframe/dumb client model, where powerful servers do most of the processing on behalf of very low-powered – and inexpensive – workstation computer terminals.) While the thin client/fat client debate rages on, TCO as a valuable cost-analysis metric has acquired significant status. While typically used to analyze the cost of workstation computers (i.e., the desktop zone), TCO methods can be applied across the network. Computer industry consulting groups like Forrester Research and the Gartner Group place the TCO cost of workstation computers at between $8,000 and $10,000. These claims accompany their contention that smart management, sound policies, and sensible procedures can bring TCO down significantly. This is a widely held view. Various TCO strategies slice the onion different ways, but common threads emerge. Following is a summary of generally accepted sound management practices for managing network costs. Standard Desktop Configuration The heart of cost containment is to be found in the desktop zone. This is where a network incurs the greatest risk of spinning out of control, because this is where users touch the network. In a 2000-workstation network, you have 2000 views of the network. While flexibility is among a computer network’s greatest strengths, uncontrolled it can evolve into a costly weakness. Where desktop configurations are idiosyncratic, the cost of reconfiguring for new or replacement workers can make TCO skyrocket. Inconsistencies across desktop platforms can make collaboration difficult, unreliable, or even impossible, impacting productivity severely. Indiscriminate software installations make a nightmare of site-license management, and makes reliable asset management nearly impossible. Clearly, no single desktop “image” suits all employees, particularly in a large company with diverse job classifications. However, it is highly likely that a fixed number of job-related desktop images will suffice – one for clerical staff, for example, another for engineers, another for managers, etc. There will always be special-case needs, of course, but even these should be handled procedurally (with a request-and-approval procedure), so that the unique installation can be recorded and tracked as part of your corporate memory. An important part of administering the desktop zone is periodically auditing the desktop image to ensure that the desktop images remain in conformance with company policy. This is important, because only with consistency at the desktop can a company enjoy TCO advantages through economies of scale – not just in procurement, but also in training and support. Enterprise-Wide Planning and Policy-based Management Consistency at the desktop is only the beginning. The hallmark of network functionality, after all, is the success with which its pieces integrate. Integration is the keyword that informed much of Chapter One. Network planning must begin at the top, with a macroscopic view of the overall network architecture and the business systems they support. This applies to planning for future enhancements, upgrades, and scalability Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 14

Woodard • Gattuccio • Brain

Networking Basics

as well. Data center, desktop, and network infrastructure must integrate fully and seamlessly, not just in the present, but in the future as well. Policy-based management is a common buzz-phrase in network management of late, but for sound reasons. If offers an a priori approach to network management, in which needs are assessed and functionality is assigned in advance of a system’s implementation. This dovetails closely with standardized desktop images, and in fact, the desktop image is an outcome of just such a policy-based management decision. The goal is constraining users to just the right palette of applications and just the right amount of computing power to allow them to complete their assigned tasks. This helps maintain consistency and scalability, reduces costs per workstation, and lightens user support costs – in part because users are not confronted with applications on which they are not trained. Perhaps nowhere is policy-based management more an issue than in the arena of online communications – everything from internal messaging to online access to the Internet. We discuss the subject of messaging policy at length in Chapter Four. Stratified Central User Support Astonishing statistics are reported about the costs of inefficient training, service and support. It has been reported, for example, that it costs upwards of $80.00 for an IS technician to make a desktop service call. It is also reported that, even on highly efficient networks (experiencing less than 5% downtime), downtime can cost a 2000-workstation network over a half-million dollars per year. These numbers may not be relevant in your company, but that is not the point. The point is, manageable and controllable factors in the arena of training, service and support can potentially shave a significant fraction from your TCO – regardless of your company’s size. User support should be centralized and stratified. Systems are best managed, and network problems and issues best tracked, when support services are administered centrally. Efficiencies of scale are achieved, service experience is better shared, and common problems are more easily identified and tracked when administered centrally. User support and problem resolution should also be stratified. By this, we mean the company’s systems and procedures should route all service calls through a common channel to a least-cost support center – typically telephone based. If unable to solve the problem, the call center should be trained to filter problems in a cost-efficient way. Reliance on an online knowledge base, FAQ (frequently asked questions) files, or online documentation library can help reduce the need for costly service calls. Only in cases of true need should you dispatch a technician to a workstation (except, of course, in cases of equipment failure). Scalable, Fault-tolerant Architecture Few businesses, if any, plan not to grow. Fortunately, modern networks, when well designed and implemented with an eye toward growth, scale easily. We’ll see in Chapters Seven and Four how network directory and communication services are highly compatible with Internet-based naming and configuration schemes, making them highly suited to virtually unlimited growth. We saw in Chapter One how client/server architecture can also add enormously to a network’s ability to scale up rapidly and easily. However, these capabilities are useful only when built in to the framework of the original network. Re-designing and reengineering a network is costly – both in technological terms and in lost productivity due to downtime. Fault-tolerance is another critical must. In Chapter Five we discuss the issue of fault-tolerance and failover capability in connection with data storage (and again in Chapter Three in connection with system administration). Fault-tolerance is a risk-reduction cost that must be weighed against lost productivity due to

Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 15

Woodard • Gattuccio • Brain

Networking Basics

downtime resulting from equipment or software system failure. Accepting the risk and the cost of expensive downtime measures poorly against the cost of robust, fault-tolerant infrastructure and equipment. Centralized Network Management and Administration Increasingly, system administration and network management tools are allowing technicians and administrators to service the network from a central point. Particularly in the area of deploying software (new or upgrades), this can offer enormous cost savings. When the deployment is for operating system software, or communication software, which may need precise configuring, the potential savings are even more significant. The arithmetic is easy. Even at just fifteen minutes per workstation, deploying an upgrade across 2000 workstations is enormously expensive – costing perhaps as much as the software itself. Sophisticated network management tools also typically include powerful network monitoring tools to allow for ongoing performance assessments and fine-tuning. These can also play a role in auditing user conformance to company policy – particularly in the area of online communication. With Internet connections increasingly common in business, user desktops extend the user’s reach globally. While a powerful tool, this may also merit attention from a management and policy-planning point of view.

Questions For Managers •

What is the population of our network? How many workstations, servers, hubs, routers, etc.? What is the geographical scope of our network? Does our infrastructure meet the demands of the network’s size and scope?

•

How do we monitor network performance, bandwidth, throughput and so on? Is our bandwidth adequate throughout the network to handle the load? How is network load measured? How is it reported? Can I see a report?

•

What is the average load on the network? What is our maximum capacity? How often do we max out? What are the implications of this? Do we lose customers? Do we lose productivity? At what rate is network traffic growing? When do we expect to overload our present capacity? Do we have the means to scale up? Is there a plan for scaling up? Can I see the plan?

•

What is our TCO? What was it last year? What is it expected to be next year (in two years, three years)? How do we calculate TCO? What variables are included? What is excluded from the calculation? How does out TCO stack up against others in our industry? How does it stack up against other businesses our size?

•

What new technologies are on the horizon that may impact our TCO in a significant way? What’s all this about thin clients and network PCs? Are they really less expensive, or do they have hidden costs? How will new networking technologies impact productivity? How will new networking technologies impact our network links with our customers and business partners? Can you show me different options with the projected total costs of each?

•

How do we handle training and user support? How much do we pay for training and support? Do we measure the effectiveness of our training and the efficiency of our support? How do we calculate the cost of these things? What are the criteria and methods? What are the results? May I see these results?

•

How do we define downtime? How do we measure downtime? What is our annualized downtime? What was it last year? Are these figures reasonable? What are the main causes of downtime? Do we have a weak link? How do we fix this?

Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 16

Woodard • Gattuccio • Brain

Networking Basics

•

Do we have standard desktop configurations? How many do we have? To what job classifications are they tied? How much does each one cost (i.e., what is the TCO on each configuration)? Are these configurations reasonable? How many exceptions are requested (and how many granted) on average?

•

Is our network fault-tolerant? To what degree? By what means do we ensure our current level of fault-tolerance? How is this measured? What does it cost? Is our present level adequate? What level of fault-tolerance is adequate? How much will it cost?

Copyright © 1998 by Interface Technologies, Inc. All rights reserved.

Page 17