Architecture (Несколько текстов для зачёта), страница 3

A Single Instruction stream, Multiple Data stream (SIMD) architecture has multiple processing elements that carry out the same instruction on separate data. For example, a SIMD machine with 100 processing elements can simultaneously multiply 100 numbers each by the number 3. SIMD processors are programmed much like SISD processors, but their operations occur on arrays of data instead of individual values. SIMD processors are therefore also known as array processors. Examples of applications that use SIMD architecture are image-enhancement processing and radar processing for air-traffic control.

A Multiple Instruction stream, Multiple Data stream (MIMD) processor has separate instructions for each stream of data. This architecture is the most flexible, but it is also the most difficult to program because it requires additional instructions to coordinate the actions of the processors. It also can simulate any of the other architectures but with less efficiency. MIMD designs are used on complex simulations, such as projecting city growth and development patterns, and in some artificial-intelligence programs.

Another factor in parallel-processing architecture is how processors communicate with each other. One approach is to let processors share a single memory and communicate by reading each other's data. This is called shared memory. In this architecture, all the data can be accessed by any processor, but care must be taken to prevent the linked processors from inadvertently overwriting each other's results.

An alternative method is to connect the processors and allow them to send messages to each other. This technique is known as message passing or distributed memory. Data are divided and stored in the memories of different processors. This makes it difficult to share information because the processors are not connected to the same memory, but it is also safer because the results cannot be overwritten.

In shared memory systems, as the number of processors increases, access to the single memory becomes difficult, and a bottleneck forms. To address this limitation, and the problem of isolated memory in distributed memory systems, distributed memory processors also can be constructed with circuitry that allows different processors to access each other's memory. This hybrid approach, known as distributed shared memory, eliminates the bottleneck and sharing problems of both architectures.

Parallel processing is more costly than serial computing because multiple processors are expensive and the speedup in computation is rarely proportional to the number of additional processors.

MIMD processors require complex programming to coordinate their actions. Finding MIMD programming errors also is complicated by time-dependent interactions between processors. For example, one processor might require the result from a second processor's memory before that processor has produced the result and put it into its memory. This results in an error that is difficult to identify.

Programs written for one parallel architecture seldom run efficiently on another. As a result, to use one program on two different parallel processors often involves a costly and time-consuming rewrite of that program.

When a parallel processor performs more than 1000 operations at a time, it is said to be massively parallel. In most cases, problems that are suited to massive parallelism involve large amounts of data, such as in weather forecasting, simulating the properties of hypothetical pharmaceuticals, and code breaking. Massively parallel processors today are large and expensive, but technology soon will permit an SIMD processor with 1024 processing elements to reside on a single integrated circuit.

Researchers are finding that the serial portions of some problems can be processed in parallel, but on different architectures. For example, 90 percent of a problem may be suited to SIMD, leaving 10 percent that appears to be serial but merely requires MIMD processing. To accommodate this finding two approaches are being explored: heterogeneous parallelism combines multiple parallel architectures, and configurable computers can change their architecture to suit each part of the problem.

In 1996 International Business Machines Corporation (IBM) challenged Garry Kasparov, the reigning world chess champion, to a chess match with a supercomputer called Deep Blue. The computer utilized 256 microprocessors in a parallel architecture to compute more than 100 million chess positions per second. Kasparov won the match with three wins, two draws, and one loss. Deep Blue was the first computer to win a game against a world champion with regulation time controls. Some experts predict these types of parallel processing machines will eventually surpass human chess playing ability, and some speculate that massive calculating power will one day substitute for intelligence. Deep Blue serves as a prototype for future computers that will be required to solve complex problems.

World Wide Web

World Wide Web (WWW), computer-based network of information resources that a user can move through by using links from one document to another. The information on the World Wide Web is spread over computers all over the world. The World Wide Web is often referred to simply as “the Web.”

Internet Topology

The Internet and the Web are each a series of interconnected computer networks. Personal computers or workstations are connected to a Local Area Network (LAN) either by a dial-up connection through a modem and standard phone line, or by being directly wired into the LAN. Other modes of data transmission that allow for connection to a network include T-1 connections and dedicated lines. Bridges and hubs link multiple networks to each other. Routers transmit data through networks and determine the best path of transmission.

The Web has become a very popular resource since it first became possible to view images and other multimedia on the Internet, a worldwide network of computers, in 1993. The Web offers a place where companies, institutions, and individuals can display information about their products, research, or their lives. Anyone with access to a computer connected to the Web can view most of that information. A small percentage of information on the Web is only accessible to subscribers or other authorized users. The Web has become a forum for many groups and a marketplace for many companies. Museums, libraries, government agencies, and schools make the Web a valuable learning and research tool by posting data and research. The Web also carries information in a wide spectrum of formats. Users can read text, view pictures, listen to sounds, and even explore interactive virtual environments on the Web.

Like all computer networks, the Web connects two types of computers–clients and servers—using a standard set of rules for communication between the computers. The server computers store the information resources that make up the Web, and Web users use client computers to access the resources. A computer-based network may be a public network—such as the worldwide Internet—or a private network, such as a company’s intranet. The Web is part of the Internet. The Internet also encompasses other methods of linking computers, such as Telnet, File Transfer Protocol, and Gopher, but the Web has quickly become the most widely used part of the Internet. It differs from the other parts of the Internet in the rules that computers use to talk to each other and in the accessibility of information other than text. It is much more difficult to view pictures or other multimedia files with methods other than the Web.

Enabling client computers to display Web pages with pictures and other media was made possible by the introduction of a type of software called a browser. Each Web document contains coded information about what is on the page, how the page should look, and to which other sites the document links. The browser on the client’s computer reads this information and uses it to display the page on the client’s screen. Almost every Web page or Web document includes links, called hyperlinks, to other Web sites. Hyperlinks are a defining feature of the Web—they allow users to travel between Web documents without following a specific order or hierarchy.

When users want to access the Web, they use the Web browser on their client computer to connect to a Web server. Client computers connect to the Web in one of two ways. Client computers with dedicated access to the Web connect directly to the Web through a router (a piece of computer hardware that determines the best way to connect client and server computers) or by being part of a larger network with a direct connection to the Web. Client computers with dial-up access to the Web connect to the Web through a modem, a hardware device that translates information from the computer into signals that can travel over telephone lines. Some modems send signals over cable television lines or special high-capacity telephone lines such as Integrated Services Digital Network (ISDN) or Asymmetric Digital Subscriber Loop (ADSL) lines. The client computer and the Web server use a set of rules for passing information back and forth. The Web browser knows another set of rules with which it can open and display information that reaches the client computer.

Web servers hold Web documents and the media associated with them. They can be ordinary personal computers, powerful mainframe computers, or anywhere in the range between the two. Client computers access information from Web servers, and any computer that a person uses to access the Web is a client, so a client could be any type of computer. The set of rules that clients and servers use to talk to each other is called a protocol. The Web, and all Internet formats, uses the protocol called TCP/IP (Transmission Control Protocol/Internet Protocol). However, each part of the Internet—such as the Web, gopher systems, and File Transfer Protocol (FTP) systems—uses a slightly different system to transfer files between clients and servers.

The address of a Web document helps the client computer find and connect to the server that holds the page. The address of a Web page is called a Uniform Resource Locator (URL). A URL is a compound code that tells the client’s browser three things: the rules the client should use to reach the site, the Internet address that uniquely designates the server, and the location within the server’s file system for a given item. An example of a URL is http://encarta.msn.com/. The first part of the URL, http://, shows that the site is on the World Wide Web. Most browsers are also capable of retrieving files with formats from other parts of the Internet, such as gopher and FTP. Other Internet formats use different codes in the first part of their URLs—for example, gopher uses gopher:// and FTP uses ftp://. The next part of the URL, encarta.msn.com, gives the name, or unique Internet address, of the server on which the Web site is stored. Some URLs specify certain directories or files, such as http://encarta.msn.com/explore/default.asp—explore is the name of the directory in which the file default.asp is found.

The Web holds information in many forms, including text, graphical images, and any type of digital media files: including video, audio, and virtual reality files. Some elements of Web pages are actually small software programs in their own right. These objects, called applets (from a small application, another name for a computer program), follow a set of instructions written by the person that programmed the applet. Applets allow users to play games on the Web, search databases, perform virtual scientific experiments, and many other actions.

The codes that tell the browser on the client computer how to display a Web document correspond to a set of rules called Hypertext Markup Language (HTML). Each Web document is written as plain text, and the instructions that tell the client computer how to present the document are contained within the document itself, encoded using special symbols called HTML tags. The browser knows how to interpret the HTML tags, so the document appears on the user’s screen as the document designer intended. In addition to HTML, some types of objects on the Web use their own coding. Applets, for example, are mini-computer programs that are written in computer programming languages such as Visual Basic and Java.

Client-server communication, URLs, and HTML allow Web sites to incorporate hyperlinks, which users can use to navigate through the Web. Hyperlinks are often phrases in the text of the Web document that link to another Web document by providing the document’s URL when the user clicks their mouse on the phrase. The client’s browser usually differentiates between hyperlinks and ordinary text by making the hyperlinks a different color or by underlining the hyperlinks. Hyperlinks allow users to jump between diverse pages on the Web in no particular order. This method of accessing information is called associative access, and scientists believe it bears a striking resemblance to the way the human brain accesses stored information. Hyperlinks make referencing information on the Web faster and easier than using most traditional printed documents.

Even though the World Wide Web is only a part of the Internet, surveys have shown that over 75 percent of Internet use is on the Web. That percentage is likely to grow in the future.

One of the most remarkable aspects of the World Wide Web is its users. They are a cross section of society. Users include students who need to find materials for a term paper, physicians who need to find out about the latest medical research, and college applicants investigating campuses or even filling out application and financial aid forms online. Other users include investors who can look up the trading history of a company’s stock and evaluate data on various commodities and mutual funds. All of this information is readily available on the Web. Users can often find graphs of a company’s financial information that show the information in several different ways.

Travelers investigating a possible trip can take virtual tours, check on airline schedules and fares, and even book a flight on the Web. Many destinations—including parks, cities, resorts, and hotels—have their own Web sites with guides and local maps. Major delivery companies also have Web sites from which customers can track their shipments, finding out where their packages are or when they were delivered.