1.3 The Benefits of Distributed Programming

Distributed programming techniques allow software to take advantage of resources located on the Internet, on corporate and organization intranets , and on networks. Distributed programming usually involves network programming in one form or another. That is, a program on one computer on a network needs some hardware or software resource that belongs to another computer either on the same network or on some remote network. Distributed programming is all about one program talking to another program over some kind of network connection, which may involve everything from modems to satellites . The distinguishing feature of distributed programs is they are broken into parts. Those parts are usually implemented as separate programs. Those programs typically execute on separate computers and the program's parts communicate with each other over a network. Distributed programming techniques provide access to resources that may be geographically distant . For example, a distributed program divided into a Web server component and a Web client component can execute on two separate computers. The Web server component can be located in Africa and the Web client component can be located in Japan. The Web client part is able to use software and hardware resources of the Web server component, although they are separated by a great distance and almost certainly located on different networks running different operating environments. Distributed programming techniques provide shared access to expensive hardware and software resources. For instance, an expensive, high-end holographic printer may have print server software that provides print services to client software. The print client software resides on one computer and the print server software resides on another computer. Only one print server is needed to serve many print clients . Distributed computing can be used for redundancy and fail over. If we divide the program up into a number of parts with each running on different computers, then we may assign some of the parts the same task. If one of the computers fails for some reason then another part of the same program executing on a different computer picks up the work. Databases can be used to hold billions, trillions, even quadrillions of pieces of information. It is simply not practical for every user to have a copy of the database. The problem is some users are located in different buildings than where the computer with the database is located. Some users are located in different cities, states, and in some instances, countries . Distributed programming techniques are used to allow users to share the massive database regardless of where they are located.

1.3.1 The Simplest Distributed Programming Models

The client-server model of distributed computing is perhaps the easiest to understand and the most commonly used. In this model, a program is divided up into two parts: One part is called the server and the other the client. The server has direct access to some hardware or software resource that the client wants to use. In most cases, the server is located on a different machine than the client. Typically, there is a many-to-one relationship between the server and the client, that is, there is usually one server fulfilling the requests of many clients. The server usually mediates access to a large database, an expensive hardware resource, or an important collection of applications. The client makes requests for data, calculations, and other types of processing. A search engine is a good example of a client-server application. Search engines are used to locate information on the Internet or on corporate and organization intranets. The client is used to obtain a keyword or phrase that the user is interested in. The client software part then passes the request to the server software part. The server has the muscle to perform the massive search for the user's keyword or phrase. The server has either direct access to the information or to other servers that have access to the information. Ideally, the server finds the keyword or phrase the user requested and returns that information to the client component. Although the client and the server are separate programs on separate computers, they make up a single application. This division of a piece of software into a client and a server is the primary method of distributed programming. The client-server model also has other forms depending on the environment. For instance, the term producer-consumer is a close cousin of client-server. Typically, client-server applications refer to larger programs and producer-consumer refers to smaller programs. Usually when the programs are at the operating system level or lower they are called producer-consumer, and when they are above the operating system level they are usually called client-server (however, there are always exceptions).

1.3.2 The Multiagent (Peer-to-Peer) Distributed Model

Although the client-server model is the most prevalent distributed programming model in use, it is not the only model. Agents are rational software components that are self directed, often autonomous, and can continuously execute. Agents can both make requests of other software components and fulfill requests of other software components . Agents can cooperate within groups to perform certain tasks collectively. In this model there is no specific client or server. The agents form a kind of peer-to-peer model where each of the components are on somewhat equal footing and each component has something to offer to the other. For example, an agent that is providing a price quote for the refurbishing of a vintage sports car might work together with other agents. Where one agent specializes in engine work, another specializes in body work, another specializes in interior design and so on. These agents may cooperatively and collectively come up with the most competitive quote for refurbishing the car. The agents are distributed because each agent is located on a different server on the Internet. The agents use an agreed-upon Internet protocol to communicate. The client-server model is a natural fit for certain types of distributed programming and the peer-to-peer agent model is a natural fit for certain types of distributed programming. We explore both types in this book. The client-server and peer-to-peer models can be used to satisfy most distributed programming demands.