A data network in a business is much like a nervous system in a living organism. Usually, when the business grows, the network also must grow to keep up with new users and newer functions. When a business suffers and shrinks, the costs associated with a larger network must be re-evaluated to determine whether they are still feasible under a smaller business organization. A network rarely goes unchanged year after year. Before you begin to write an upgrade plan for part or all of a network, you first must determine that there is a need for an upgrade. There are several reasons to upgrade a network:
The first of these reasons, user complaints, probably will never go away. No matter how fast the network or how powerful the machine, there will always be someone who wants more. Usually a network administrator has a good overview of the people or departments that are major consumers of network resources and can filter out unrealistic complaints. So when deciding when and whether you need to perform some kind of overhauling of the network, you should carefully research your current network's capacity and compare it to the business needs currently loading the network. Sometimes a simple overview of the network can reveal that all you must do to satisfy one or two small bottlenecks is reconfigure part of the network. Usually a reconfiguration is cheaper and easier to implement than a major upgrade. It's much easier to upgrade a single, growing department to a faster switch than it is to overhaul the entire network. When reconfiguring the network will not solve capacity problems, it might be time to look at other media to handle the network traffic. Chapter 6, "Wiring the NetworkCables, Connectors, Concentrators, and Other Network Components," gives a good overview of the specifications and capabilities of this most basic part of the network. Whether you are planning an upgrade or simply extending your network into a larger geographic space, it's not necessarily a good idea to shop around and settle for the cheapest solution. Installing cabling in ceilings, walls, and floors is a labor-intensive, expensive item in an upgrade budget. Yet if you spend the money up front to install good-quality cables, connectors, and the like, you will probably quickly recoup the cost in a few years when upgrading to even faster networking technologies. For example, you can use Category 5, 5e or 6 for a 10/100 Ethernet network. Any of these cable types can also support Gigabit Ethernet. However, Category 6 cable can support faster transmission speeds and provide greater signal quality, making it a good candidate for even faster network standards. If you want to make a long-term investment in your UTP cable plant, use Category 6 rather than Category 5 or 5e. Tip When you get around to pulling new cables, don't think just one generation ahead. Pulling cables, connecting them to the user's faceplate at the terminal point, and connecting them to patch panels or other devices in the wiring closets can often be the most expensive component of your network. It's easy to replace a workstation. It's somewhat more expensive to replace a server. Pulling cables is very expensive, and it can involve replacing not just the cables, but also the connectors (at both ends of the connection). It can also involve replacing network adapter cards at both ends. If you are replacing cables today, consider fiber-optic cabling for backbone connections, as well as connections between wiring closets. In a few years the same may even apply to connections from the wiring closet to the desktop. It's all based on applications, and the bandwidth they require is constantly growing. As an alternative to pulling cables for lower-capacity network segments, consider using wireless Ethernet. IEEE 802.11g support is widespread in business-class notebook computers, and you can buy WAPs that include 10/100 Ethernet switches to enable a mixed wired/wireless network segment. Keep in mind that you should use the strongest possible wireless security settings, such as WPA2, to protect your wireless network from unauthorized use. There are other chapters you might want to consult when planning an upgrade. Chapter 7, "Network Interface Cards," covers the basics of this topic. You also can find information there about some of the more recent functionality that has been incorporated into NICs, such as Wake on LAN and server NIC load balancing. After you have the wiring and NIC issues settled, Chapter 8, "Network Switches," guides you through the process of replacing overloaded department hubs with switches to improve performance and connect dissimilar LAN segments. If your network still implements generic hubs one of the first upgrades you should consider is replacing them with switches. Ethernet hubs are now legacy equipment. If you are considering wireless Ethernet, see Chapters 1823 for information about technologies, speeds, and security features. Adopting new technology usually entails additional tasks such as training users and administrative personnel in the use or management of new products. For example, a major paradigm shift, such as migrating an all-Novell network to Windows XP/Vista clients and Windows Server 2003 servers, would have to include months of training for network administrative personnel. For all practical purposes, TCP/IP is now the de facto networking standard for LANs. If you are still running an AppleTalk or IPX/SPX network, you might want to think about future support for these products. Novell already offers TCP/IP support, and there are products that allow interaction between Novell Directory Services (NDS) and Windows Active Directory. Although NDS still might be a useful tool, IPX/SPX LANs are good candidates for upgrading to TCP/IP, especially if your business connects to the Internet. Tip Although Novell's NetWare 6.5 has brought a lot of raves from its implementers, it is my considered opinion that the traditional capabilities of NetWare are a thing of the past. There are many directory services (most notably Microsoft's Active Directory) that make this major capability now a minor concern. Novell's purchase of SuSE Linux in 2003 and its subsequent incorporation of SuSE Linux and NetWare 6.5 into a single integrated product known as Novell Open Enterprise Server suggest that Novell sees its future as being one that embraces open network standards such as Linux. It also enhances those standards with products such as ZENworks, DirXML, iPrint, iFolder, and others. Another technology that might be ready for a change is the Token-Ring LAN. Although Token-Ring technology is a great LAN solution, the fastest Token-Ring adapters run at 100Mbps, and most run at 4 or 16Mbps. These speeds are outstripped by full-duplex Fast Ethernet (200Mbps effective speed) and Gigabit Ethernet. And when it comes to the budget, the number of manufacturers of Ethernet equipment dwarfs those who make Token-Ring hardware. Because of the economics of scale, and competition, Ethernet is a far cheaper solution in virtually every case. You probably won't even need an adapter for your clients: 10/100 (and sometimes Gigabit) Ethernet adapters are built into almost all recent PCs. For newcomers, Chapter 13, "Ethernet: The Universal Standard," and Chapter 24, "Overview of the TCP/IP Protocol Suite," will get you started on the planning process. Other chapters can fill in the details of some aspects typical of TCP/IP-based networks. For example, Chapter 25, "Basic TCP/IP Services and Applications," will help you determine which of the basic TCP/IP-based applications will be useful in your network. Chapter 28, "BOOTP and Dynamic Host Configuration Protocol (DHCP)," and Chapter 29, "Network Name Resolution," cover the basics of automating network addressing and name resolution. Finally, one "upgrade" that every network administrator should plan for, if it has not happened yet, is to provide a firewall between the company network and any outside network connection. Chapter 45, "Firewalls," covers this area. The chapters immediately preceding cover other aspects of network and systems security. Today a firewall is not a luxury. It is a necessity. It doesn't matter if you are connecting different LAN segments within a company intranet, or if you are making a connection to the Internet. A good firewall is now something you cannot ignore. Even SOHO users should use some type of firewall to protect themselves from the malicious problems that can result from a wide area connection. For a corporate connection to the Internet, a good firewall, no matter what the cost, is an absolute necessity. And in some professions, such as the pharmaceutical industry, or for those corporations that do business with the government, it is required. Most SOHO users will receive adequate protection from a software firewall (such as the firewall integrated into Windows XP Service Pack 2 or third-party products such as Zone Alarm), but corporate servers and networks need hardware firewall appliances for better protection. Both types of firewalls are discussed in Chapter 45. Determining User Requirements and ExpectationsOne of the simplest methods of finding out what is needed on the network is to ask the users. Although this technique might not give you the most accurate results, it will at least give you an idea of what the user community expects from the network. Conducting a simple written survey can bring light to factors that administrative and support personnel might not be aware of. Similar to surveying end users is soliciting suggestions from support staff who encounter user problems daily. Logs of support calls can be a valuable source of information. If users are making the same mistakes over and over again, then training, not upgrading, usually can solve the problem. When examining help-desk logs, don't limit your scope to the problems. Carefully examine what was done to solve the problems. Maybe your help-desk personnel aren't giving out the best advice. However, the most basic way to determine overall capacity needs is to establish baseline data for your network components and then make comparisons on a regular basis with the production network. By keeping data that reflects the baseline mode of operation, you will have empirical data that can be used to make projections about future use. By regularly benchmarking your systems and keeping track of the data, you also can become aware of capacity problems that begin to creep up before your projections or expectations. Most operating systems provide for monitoring of a sort. Windows 2000 Performance Monitor, or Windows 2003/XP's System Monitor, for example, can give you very detailed statistics on a server's performance, including network performance. Chapter 5, "Protecting the Network: Preventative Maintenance Techniques," gives an overview of this tool. In larger networks, using a management console that understands SNMP and RMON can help you baseline and monitor many different components in the network. Analysis software can help prepare reports that can be used to justify network changes. Both SNMP and RMON are examined in Chapter 49, "Network Testing and Analysis Tools." Other factors that can be reviewed to determine whether changes are merited include the following:
Maintaining Support for Legacy ApplicationsThere comes a time in every application's life when it really should just die and go away. However, when you consider the costs associated with replacing a legacy software application with one that is state-of-the-art, sometimes what you find might lead you to listen to users who like their software and want to keep the application around a little longer. The following are some hidden costs you might overlook at first glance:
Maintenance and support costs can be hidden in part of the budget for the department that an application or a hardware platform supports. When multiple suppliers are involved, you will usually find multiple contracts, some of which even overlap each other. Another problem with ongoing support costs is that manufacturers raise these costs when products become outdated to help encourage users to adopt newer ones. The number of employees you dedicate to a particular part of the network is an important cost. If you have a large staff whose responsibility is mostly maintaining an old application, consider the costs associated with them, from salary to overhead, and decide whether it might be better spent training them on newer technology and replacing the legacy system. Legacy applications are usually either loved or hated by their users. Some people think with the mindset "but we've always done it that way" and do not want to change for fear of the unknown and love of the familiar. Other users who are used to better technology might hate a legacy application. The point to remember here is that it does not always matter what the user thinks about the application. What is most important to the business's bottom line is how the costs associated with the change compare with the benefits the company receives. |