Now that you've had a chance to explore many of the subsystems of modern serversincluding processors, motherboards, buses, and network interfacesit's time to get your server(s) deployed. With what you have learned already, it should be possible to make a good selection of server type. However, knowing what server you want and choosing the right server for the job at hand is something of an art form. That selection entails choosing a server that allows you to service the number of clients you have, not way too many and certainly not too fewsomething that is often referred to as right sizing or capacity planning. This chapter describes some of the considerations involved in trying to size your systems correctly, as well as some of the tools available to help. When you select a server to deploy, you essentially choose a platform to build on. It's quite possible that you will buy a specialized server designed for a purpose such as a network attached storage (NAS), deploy that server, and then never change that server's hardware or base network operating system in any significant way. You might buy a server with a contract, and in this case, whenever something goes wrong or you need to beef up the system, your value-added reseller (VAR) sends over Vinny, who keeps your server in line or gives it a shot of what it needs. It's just as common to deploy a general-purpose server and evolve that system over time. If you go this route, you need to look at your deployment with an eye to the future. For a general-purpose serveror any other serveryou need an upgrade path if you intend to keep that server in service for an extended period. The period that a server can be in service translates directly to your organization's return on investment and to its profit. So there is a strong incentive to make some good choices that will give your server added life. This topic is discussed further in this chapter. In the past, few people built their own servers. Most chose to purchase a server from a large original equipment manufacturer (OEM) or from a company that built no-named servers, called white boxes. As more and more server technologies have become mainstream, performance desktop computers have begun to feature many of the same technologies in their builds that servers traditionally had. It has become much easier for a person with basic computer-building skills to construct a custom-built server from standardized parts. You might not want to build an enterprise-class eight-way SMP system from scratch, but a dual- or even a quad-processor system is within the capabilities of most people who read books like this one. To successfully build a server, you need to create balance between your expectations of the server's purpose and the components necessary to achieve that purpose. You shouldn't expect to save money by choosing to build your own server. The primary benefits of building your own server don't lie in cost, but rather in the customization and upgrade potential of a home-built server, as well as the ability to service it yourself. We'll describe some of the pluses and minuses of building a server and whether you should attempt to build your own. Finally, you need to take into account a number of other considerations when you deploy a server. Servers have different licensing requirements from desktop computers, as do the applications running on them. You need to know whether your quad processor needs one, two, or even four operating system licenses. At the moment, it appears that vendors are only requiring that a dual-core processor have only a single license for software and hardware. Some software products have licenses that don't use processor counts at all; instead, their licenses are based on the number of connections or some other factors. Some licenses can be mixed and matched. Licenses are a recurring nightmare for any but the smallest of shops. You are definitely not in Kansas anymore! This chapter addresses how you can comply with license terms and discusses management software to help you with all this complexity. |