Network Design and Management

The systems development life cycle (SDLC) is a structured approach to the development of a business system. This approach often includes planning, analysis, design, implementation, and support. Although virtually every company that uses SDLC and every textbook that teaches SDLC has its own slightly different variation of the methodology, most agree that the SDLC includes the following phases:

• Planning—Identify problems, opportunities, and objectives.
• Analysis—Determine information requirements, analyze system needs, and prepare a written systems proposal.
• Design—Design and build the system recommended at the end of the analysis phase and create the documentation to accompany the system.
• Implementation—Install the system and prepare to move from the old system to the new system; train the users.
• Maintenance—Correct and update the installed system as necessary.

The idea of phases is critical to the SDLC concept. The intent of SDLC is for phases not to be disjointed steps in a big plan, but overlapping layers of activity. A second critical concept is that of the cycle. After a system has been maintained for a period of time, it is relatively common to restart the planning phase— hence, another cycle—in an attempt to seek a better solution to the problem.

One technique used to model a corporation’s network environment is to create connectivity maps. More precisely, three different modeling techniques can be used, depending on what type of network you are modeling: wide area connectivity maps, metropolitan area connectivity maps, and local area connectivity maps.

In order to create a wide area connectivity map, the modeler begins by identifying each site or location in which the company has an office. Each fixed site is denoted by a circle; mobile or wireless sites are indicated by circles containing the letter M; and external sites, such as suppliers or external agents, are denoted by circles containing the letter E. A solid line between two sites indicates a desired path for data (or voice) transmission. If a company desires a metropolitan area network connection between one of its offices and another business, such as an Internet service provider (ISP), it can use a metropolitan area connectivity map to outline this connection and define the desired network characteristics. To examine the nodes in a wide area connectivity map in more detail, an analyst can expand each individual site into a local area connectivity map. The local area network design can then be performed in one or two stages, depending upon the level of detail desired. If only an overview of a local network is desired, then the analyst can create a local area overview connectivity map. If more detail is desired, the analyst can create a local area detailed connectivity map.

Analyzing and designing a new computer system can be time-consuming and expensive. While the project is in the analysis phase, and before a system is designed and installed, a feasible solution must be found. The term “feasible” has several meanings when it’s applied to computer-based projects. The technical feasibility of a system is the extent to which the system can be created and implemented using currently existing technology. A system’s financial feasibility is the extent to which the system can be created, given the company’s current finances. When a system demonstrates operational feasibility, it operates as designed and implemented. A system’s time feasibility is the extent to which the system can be installed in a timely fashion and meets organizational needs.

To determine the cost of a system, it is necessary to include all possible costs. But to get a comprehensive understanding of the cost of the system, you must also calculate the recurring costs of the proposed system. Once the one-time and recurring costs have been established, it is time to determine the benefits that will result from the proposed system. When calculating benefits, you will need to include both tangible benefits and intangible benefits. Now that the costs and benefits have been determined, you can apply them to a payback analysis. When performing a payback analysis calculation, you should show all dollar amounts using the time value of money. This means that if something is going to cost one dollar one year from now, you need to put away less than a dollar today to pay for it. This also means that if something is going to cost one dollar one year from now, you need to put away less than a dollar today to pay for it. A payback analysis helps you to determine the approximate time for a project payback, or return on investment (ROI), to occur.

If you design a system for a company, and the system is not capable of supporting the traffic generated within the company, response times will be sluggish, and users may not be able to complete their work on time. Capacity planning involves trying to determine the amount of network bandwidth necessary to support an application or a set of applications. Capacity planning is a fairly difficult and time-consuming operation. A number of techniques exist for performing capacity planning, including linear projection, computer simulation, benchmarking, and analytical modeling.

Linear projection involves predicting one or more network capacities based on the current network parameters and multiplying those capacities by some constant. Some systems, however, do not follow a linear projection. If you apply a linear projection to these systems, you may produce inaccurate predictions. In these cases, an alternate strategy is required. A computer simulation involves modeling an existing system or a proposed system using a computer-based simulation tool and subjecting the model to varying degrees of user demand (called load). Computer simulations are difficult to create, mainly because it is easy to make mistakes in the modeling process and difficult to discover them. Benchmarking involves generating system statistics under a controlled environment and then comparing those statistics against known measurements. Unfortunately, like simulation, this process can also suffer from possible errors. Analytical modeling involves the creation of mathematical equations to calculate various network values.

Creating a baseline for an existing computer network involves measuring and recording a network’s state of operation over a given period of time. Creating a baseline actually involves capturing many network measurements over all segments of a network, including numerous measurements on workstations, user applications, bridges, routers, and switches. Baseline studies can be started at any time but are most effective when they are initiated during a time when the network is not experiencing severe problems, such as a node failure or a jabber. Therefore, before you begin a baseline study, you must extinguish all immediate fires and try to get the network into fairly normal operation. Because you will be generating a large number of statistics, you will want to have access to a good database or spreadsheet application to keep the data organized. Once the database or spreadsheet has been set up, you are ready to begin your baseline study.

The next question is, on what items are you going to collect baseline information? You may find it useful to collect information on items such as system users, system nodes, operational protocols, network applications, and network utilization levels. Once you have collected and analyzed network utilization data, you can make several important observations. First, you can detect when a network may be reaching saturation. A second observation you can make is when peak periods of network use occur. Making observations about peak periods of network use is easiest when you graph network activity data. Examining the amount of traffic on each node also yields valuable information about network performance. Once you have performed the baseline study, don’t stop observing the network. For a baseline study to be really effective, you need to maintain it.

Once the analysis and design phases of network development are completed and the computer network is in place and operating, it is the network administrator’s responsibility to keep it running. Keeping a network running involves making repairs on failed components, installing new applications and updating the existing ones, keeping the system’s existing users up to date, and looking for new ways to improve the overall system and service level.

Because many network administrators are dealing with both computers and people, they need the skills necessary to work with both. A checklist of skills for the network administrator would include a wide platform of technology skills, including, but not limited to, knowledge of local area networks, wide area networks, voice telecommunications systems, data transmission systems, video transmission, basic hardware concepts, and basic software skills. A network administrator should also have interpersonal skills, including the ability to talk to users in order to service problems and explore new applications. Along with interpersonal skills, a network administrator also needs training skills, which involve the ability to train users or other network support personnel. To make effective use of limited resources, a network administrator should also possess a number of common management skills, including budget management, statistical, time management, project management, and policy creation and enforcement skills. To learn new skills and demonstrate proficiency within a particular area, the network administrator can obtain certification.

Computer networks are in a constant state of change. To support changes to a network, a network administrator needs funding. If properly generated, statistics can be used to support the request for a new system or modifications to an existing system. Four statistics, or measures, that are useful in evaluating networks are mean time between failures, mean time to repair, availability, and reliability. Mean time between failures (MTBF) is the average time a device or system will operate before it will fail. Mean time to repair (MTTR) is the average time necessary to repair a failure within the computer network. This time includes the time necessary to isolate the failure. The third statistic, availability, is the probability that a particular component or system will be available during a fixed time period. A component or network with a high availability (near 1.0) is almost always operational. For simplicity, however, you can calculate availability by simply subtracting the downtime from the total available time and then dividing by the total available time. The fourth statistic, reliability, calculates the probability that a component or system will be operational for the duration of a transaction of time t. Reliability is defined by the equation: R(t) = e^(–bt). A reliability of exactly 1.0 means the network or device is reliable 100 percent of the time.

The two categories of diagnostic tools are the tools that test and debug the network hardware, and the tools that analyze the data transmitted over the network. Finally, the command center and the help desk should be considered.

Tools that test and debug network hardware range from very simple devices to more elaborate, complex devices. Three common testing devices are electrical testers (the simplest), cable testers, and local area network testers (the most elaborate). An electrical tester will show if voltage is on a line, and if so, how much voltage. Cable testers are slightly more elaborate devices. They can verify connectivity and test for line faults, such as open circuits, short circuits, reversed circuits, and crossed circuits. Local area network testers can operate on Ethernet and token ring networks, whether they have switches or not. Some local area network testers have a display that graphically shows a network segment and all of the devices attached to it.

The second category of diagnostic tools covers tools that analyze data transmitted over the network. These tools include protocol analyzers and devices or software that emulate protocols and applications. One of the most common of these tools is the traffic analyzer or protocol analyzer. Each packet’s protocol is analyzed, and statistics are generated that show which devices are talking to each other and which applications are being used. This information can then be used to update the network, so that it operates more effectively.

To assist network administrators and information technologists in doing their jobs, businesses have control centers for their computing services. It contains, in one easily accessible place, all the network documentation, including network resource manuals, training manuals, baseline studies, all equipment documentation, user manuals, vendor names and telephone numbers, procedure manuals, and forms necessary to request services or equipment. The control center can also contain a training center to assist users and other information technologists.

One of the more important elements of a control center is the help desk. Whether it’s called upon to address hardware problems, answer questions about running a particular software package, or introduce the company’s users to new computing services, the help desk is the gateway between the user and computing and network services. When users know a friendly person is available to turn to for any computing problems, much less computer system and computer user friction exists.

A network management protocol facilitates the exchange of management information between network devices. This information can be used to monitor network performance, find network problems, and then solve those problems—all without having any network personnel physically touch the affected device. Simple Network Management Protocol (SNMP) is an industry standard designed originally to manage Internet components; it is now also used to manage wide area network and telecommunications systems.

SNMP is based on the following set of principles. Network objects consist of network elements such as servers, mainframe computers, printers, hubs, bridges, routers, and switches. Each of these elements can be classified as either managed or unmanaged. A managed element has management software, called an agent, running in it and is more elaborate and expensive than an unmanaged element. A second type of object—the SNMP manager software—controls the operations of a managed element and maintains a database of information about all managed elements. The database that holds the information about all managed devices is called the Management Information Base (MIB). The information stored in the MIB can be used to repair or manage the network, or simply to observe the operation of the network. Managed elements are monitored and controlled using three basic SNMP commands: read, write, and trap. The read command is issued by a manager to retrieve information from the agent in a managed element. The write command is also issued by a manager but is used to control the agent in a managed element. By using the write command, a manager can change the settings in an agent, thus making the managed element perform differently.

More often than not, the SNMP manager requests information directly from a managed element on the same network. But what if a manager wants to collect information from a remote network? Remote Network Monitoring (RMON) is a protocol that allows a network administrator to monitor, analyze, and troubleshoot a group of remotely managed elements. RMON is defined as an extension of SNMP. RMON can be supported by hardware monitoring devices, through software, or through a combination of hardware and software. RMON can collect several basic kinds of information, such as number of packets sent, number of bytes sent, number of packets dropped, host statistics, and certain kinds of events that have occurred. A network administrator can find out how much bandwidth or traffic each user is imposing on the network and can set alarms in order to be alerted of impending problems.