Information Technology Concepts

Transcription

1 PART 2 Information Technology Concepts Chapter 3 Chapter 4 Chapter 5 Chapter 6 Chapter 7 Hardware: Input, Processing, and Output Devices Software: Systems and Application Software Organzing Data and Information Telecommunications and Networks The Internet, Intranets, and Extranets

2 CHAPTER 3 Hardware: Input, Processing, and Output Devices PRINCIPLES LEARNING OBJECTIVES Assembling an effective, efficient computer system requires an understanding of its relationship to the information system and the organization. The computer system objectives are subordinate to, but supportive of, the information system and the needs of the organization. When selecting computer devices, you also must consider the current and future needs of the information system and the organization. Your choice of a particular computer system device should always allow for later improvements. Describe how to select and organize computer system components to support information system (IS) objectives and business organization needs. Describe the power, speed, and capacity of central processing and memory devices. Describe the access methods, capacity, and portability of secondary storage devices. Discuss the speed, functionality, and importance of input and output devices. Identify popular classes of computer systems and discuss the role of each.

3 Hardware: Input, Processing, and Output Devices Chapter 3 89 INFORMATION SYSTEMS IN THE GLOBAL ECONOMY PORSCHE AG, GERMANY Auto Manufacturer Upgrades Its Computer Hardware Porsche is known for manufacturing world-class sports cars, among them its famous 911 and Boxster automobiles. The company built on its reputation when it recently introduced the four-wheel-drive Cayenne, the world s fastest SUV, and the Carerra GT, a super sports car with a powerful V10 engine. Porsche s headquarters are in Stuttgart, Germany; it maintains production facilities in Stuttgart and Leipzig, and it has dealerships and sales offices worldwide. Annual sales are at record levels approaching a jaw-dropping $7 billion and its employees number just under 10,000. To maintain its market leadership, Porsche must employ information systems as cutting edge as its cars. The company was one of the first auto makers to introduce enterprise resource planning (ERP) software to support its accounting, finance, purchasing, and material-management business processes. It implemented its ERP system in the early 1990s and chose then state-of-the-art Hewlett-Packard V-class servers as the underlying computer hardware to handle some 1.5 million daily transactions quickly and reliably. Over the years, however, the volume of transactions its system must handle has more than doubled. This increase comes from four factors: (1) Porsche plans to use new ERP software modules to support additional business processes, (2) the company introduced new automobile models, (3) it built a new production plant for the Cayenne and Carerra GT, and (4) the number of users of the ERP system exceeded 4,000. Porsche Information Kommunikation Services (PIKS) GmbH is a wholly owned subsidiary of Porsche AG, with headquarters in Stuttgart. Its 86 employees are responsible for planning and operating the IS infrastructure for the entire Porsche group, including networks, servers, security systems, and storage systems. For over a year, PIKS evaluated several different computer manufacturers and hardware options to meet the new processing requirements. It was critical that the new hardware not increase the company s hardware budget. Obviously, the new hardware must work well with existing components of the infrastructure (software, network, and other computer hardware). Importantly, the new hardware must be extremely reliable and available. To meet the company s computing needs, PIKS decided to replace its HP V- class servers with two HP Superdome servers, each with 24 processors and 28 gigabytes of RAM. This hardware upgrade also provides for increased processing power to meet future needs; the PA-8700 processors in the current version of the Superdome server can be upgraded to processors from the Intel Itanium processor family to double the processing capability. As you read this chapter, consider the following: How are companies using computer hardware to compete and meet their business objectives? How do organizations go about selecting computer hardware and what must you know to assist in this process?

4 90 Part 2 Information Technology Concepts Why Learn About Hardware? Organizations invest in computer hardware to improve worker productivity, increase revenue, reduce costs, and provide better customer service. Those that don t may be stuck with outdated hardware that often fails and cannot take advantage of the latest software advances. As a result, obsolete hardware can place an organization at a competitive disadvantage. Managers, no matter what their career field and educational background, are expected to know enough about hardware to ask tough questions to invest wisely for their area of the business. Managers in marketing, sales, and human resources often help IS specialists assess opportunities to apply computer hardware and evaluate the options and features specified for the hardware. Managers in finance and accounting especially must also keep an eye on the bottom line, guarding against overspending, yet be willing to invest in computer hardware when and where business conditions warrant it. hardware Any machinery (most of which use digital circuits) that assists in the input, processing, storage, and output activities of an information system. Today s use of technology is practical intended to yield real business benefits, as seen with Porsche. Employing information technology and providing additional processing capabilities can increase employee productivity, expand business opportunities, and allow for more flexibility. As we already discussed, a computer-based information system (CBIS) is a combination of hardware, software, database(s), telecommunications, people, and procedures all organized to input, process, and output data and information. In this chapter, we concentrate on the hardware component of a CBIS. Hardware consists of any machinery (most of which use digital circuits) that assists in the input, processing, storage, and output activities of an information system. The overriding consideration in making hardware decisions in a business should be how hardware can be used to support the objectives of the information system and the goals of the organization. COMPUTER SYSTEMS: INTEGRATING THE POWER OF TECHNOLOGY A computer system is a special subsystem of an organization s overall information system. It is an integrated assembly of devices centered on at least one processing mechanism utilizing digital electronics that are used to input, process, store, and output data and information. Putting together a complete computer system, however, is more involved than just connecting computer devices. In an effective and efficient system, components are selected and organized with an understanding of the inherent trade-offs between overall system performance and cost, control, and complexity. For instance, in building a car, manufacturers try to match the intended use of the vehicle to its components. Racing cars, for example, require special types of engines, transmissions, and tires. The selection of a transmission for a racing car, then, requires not only consideration of how much of the engine s power can be delivered to the wheels (efficiency and effectiveness) but also how expensive the transmission is (cost), how reliable it is (control), and how many gears it has (complexity). Similarly, organizations assemble computer systems so that they are effective, efficient, and well suited to the tasks that need to be performed. As we saw in the opening vignette, people involved in selecting their organization s computer hardware must have a clear understanding of the business requirements so they can make good acquisition decisions. Here are several examples of applying business understanding to reach critical hardware decisions. ARZ Allgemeines Rechenzentrum GmbH (ARZ) is one of Austria s leading providers of information services, processing more than 8 million transactions per day for financial and medical institutions. ARZ is keenly interested in reducing processing costs, maintaining

5 Hardware: Input, Processing, and Output Devices Chapter 3 91 sufficient capacity to handle an increasing workload, and providing highly reliable processing. Peter Gschirr, information technology director at ARZ, selected two large, extremely powerful IBM zseries mainframe computers to handle the workload. 1 As a result of declining ticket sales, Continental Airlines management mandated that all system development projects must pay back their costs in less than one year. In addition, the Continental Technology Group needed a safe, reliable, and user-friendly development environment to write new application software. Jack Wang, managing director of the Technology Group, acquired a Hewlett- Packard HP NonStop S8600 server to run customer-service applications and a Non- Stop S74000 server for software development and backup in the event the primary server fails. These high-end, fault-tolerant servers are frequently used for critical financial transactions in the financial services and electronic commerce industries. 2 Air Products and Chemicals, Inc., is an international supplier of industrial gases and related equipment, as well as specialty chemicals. The company employs unique software that requires powerful, high-performance computers to conduct simulations and computations. Eric Werley of Air Products technical computing group recognized several shortcomings in using servers from three different manufacturers: The acquisition and ongoing maintenance of these server systems were costly, and the servers were difficult to manage due to lack of standardization. Air Products migrated to Dell computer hardware, including powerful Dell Dual Precision 410 and Dell GxPro workstations and Dell PowerEdge servers to eliminate these problems. 3 Specialized Bicycles is a pioneer in designing and manufacturing high-performance bicycles, helmets, and other cycling accessories. When Specialized Bicycles wanted to offer its products over the Web, it selected computer hardware that could rapidly increase computing capacity, provide high reliability so that the Web site was always available, and easily be integrated with the rest of the organization s hardware. To meet these requirements, Ron Pollard, chief information officer, went with Sun Microsystems Enterprise 450 Server. 4 Sullivan Street Bakery, started in 1994 by an art student and anthropologist, is today one of the most popular Italian bakeries in New York. The owners chose Apple s Mac computers because they are efficient and easy to use. The Macs make managing production almost effortless, and they support billing and inventory control. Workers don t have to worry about how to use the computers and can concentrate instead on maintaining the quality of their hand-crafted traditional Italian-style breads. 5 As each of these examples demonstrates, assembling the right computer hardware requires an understanding of its relationship to the information system and the needs of the organization. Remember that the computer hardware objectives are subordinate to, but supportive of, the information system and the needs of the organization. The components of all information systems such as hardware devices, people, and procedures are interdependent. Because the performance of one system affects the others, all of these systems should be measured according to the same standards of effectiveness and efficiency, given the constraints of cost, control, and complexity. When selecting computer hardware, you also must consider the current and future uses to which these systems will be put. Your choice of a particular computer system should always allow for later improvements in the overall information system. Reasoned forethought a Porsche needed cost-effective, reliable, powerful computers on which to run its ERP software. Source: Getty Images.)

6 92 Part 2 Information Technology Concepts trait required for dealing with computer, information, and organizational systems of all sizes is the hallmark of a true business professional. central processing unit (CPU) The part of the computer that consists of three associated elements: the arithmetic/logic unit, the control unit, and the register areas. arithmetic/logic unit (ALU) Portion of the CPU that performs mathematical calculations and makes logical comparisons. Hardware Components Computer system hardware components include devices that perform the functions of input, processing, data storage, and output (see Figure 3.1). To understand how these hardware devices work together, consider an analogy from a paper-based office environment. Imagine a one-room office occupied by a single individual. The human being (the processor) is capable of organizing and manipulating data. The person s mind (register storage) and the desk occupied by the human being (primary storage) are places to temporarily store data. Filing cabinets fill the need for a more permanent form of storage (secondary storage). In this analogy, the incoming and outgoing mail trays can be understood as sources of new data (input) or as places to put the processed paperwork (output). Communications Devices Figure 3.1 Computer System Components These components include input devices, output devices, communications devices, primary and secondary storage devices, and the central processing unit (CPU). The control unit, the arithmetic/ logic unit (ALU), and the register storage areas constitute the CPU. Input Devices Processing Device Control Unit Register Storage Area Memory (Primary Storage) Secondary Storage Arithmetic/ Logic Unit Output Devices control unit Part of the CPU that sequentially accesses program instructions, decodes them, and coordinates the flow of data in and out of the ALU, the registers, primary storage, and even secondary storage and various output devices. register High-speed storage area in the CPU used to temporarily hold small units of program instructions and data immediately before, during, and after execution by the CPU. primary storage (main memory; memory) Part of the computer that holds program instructions and data. The ability to process (organize and manipulate) data is a critical aspect of a computer system, in which processing is accomplished by an interplay between one or more of the central processing units and primary storage. Each central processing unit (CPU) consists of three associated elements: the arithmetic/logic unit, the control unit, and the register areas. The arithmetic/logic unit (ALU) performs mathematical calculations and makes logical comparisons. The control unit sequentially accesses program instructions, decodes them, and coordinates the flow of data in and out of the ALU, the registers, primary storage, and even secondary storage and various output devices. Registers are high-speed storage areas used to temporarily hold small units of program instructions and data immediately before, during, and after execution by the CPU. Primary storage, also called main memory or just memory, is closely associated with the CPU. Memory holds program instructions and data immediately before or immediately after the registers. To understand the function of processing and the interplay between the CPU and memory, let s examine the way a typical computer executes a program instruction. Hardware Components in Action The execution of any machine-level instruction involves two phases: the instruction phase and the execution phase. During the instruction phase, the following takes place: Step 1: Fetch instruction. The fetch stage reads a program s instructions and any necessary data into the processor. Step 2: Decode instruction. The instruction is decoded and is passed to the appropriate processor execution unit. There are several execution units: The arithmetic/logic unit

7 Hardware: Input, Processing, and Output Devices Chapter 3 93 performs all arithmetic operations, the floating-point unit deals with noninteger operations, the load/store unit manages the instructions that read or write to memory, the branch processing unit predicts the outcome of a branch instruction in an attempt to reduce disruptions in the flow of instructions and data into the processor, the memorymanagement unit translates an application s addresses into physical memory addresses, and the vector processing unit handles vector-based instructions that accelerate graphics operations. Steps 1 and 2 are called the instruction phase, and the time it takes to perform this phase is called the instruction time (I-time). The second phase is the execution phase. During the execution phase, the following steps are performed: Step 3: Execute the instruction. The execution stage is where the hardware element, now freshly fed with an instruction and data, carries out the instruction. This could involve making an arithmetic computation, logical comparison, bit shift, or vector operation. Step 4: Store results. During this step, the results are stored in registers or memory. Steps 3 and 4 are called the execution phase. The time it takes to complete the execution phase is called the execution time (E-time). After both phases have been completed for one instruction, they are again performed for the second instruction, and so on. The instruction phase followed by the execution phase is called a machine cycle (see Figure 3.2). Some processing units can speed up processing by using pipelining, whereby the processing unit gets one instruction, decodes another, and executes a third at the same time. The Pentium 4 processor, for example, uses two execution unit pipelines. This gives the processing unit the ability to execute two instructions in a single machine cycle. instruction time (I-time) The time it takes to perform the fetch-instruction and decodeinstruction steps of the instruction phase. execution time (E-time) The time it takes to execute an instruction and store the results. machine cycle The instruction phase followed by the execution phase. pipelining A form of CPU operation in which there are multiple execution phases in a single machine cycle. Processing Device Control Unit ALU (2) Decode (3) Execute I-Time E-Time (1) Fetch Registers (4) Store Memory Figure 3.2 Execution of an Instruction In the instruction phase, a program s instructions and any necessary data are read into the processor (1). Then the instruction is decoded so the central processor can understand what is to be done (2). In the execution phase, the ALU does what it is instructed to do, making either an arithmetic computation or a logical comparison (3). Then the results are stored in the registers or in memory (4). The instruction and execution phases together make up one machine cycle. PROCESSING AND MEMORY DEVICES: POWER, SPEED, AND CAPACITY The components responsible for processing the CPU and memory are housed together in the same box or cabinet, called the system unit. All other computer system devices, such as the monitor and keyboard, are linked either directly or indirectly into the system unit housing. As discussed previously, achieving IS objectives and organizational goals should be the primary consideration in selecting processing and memory devices. In this section, we investigate the characteristics of these important devices.

8 94 Part 2 Information Technology Concepts MIPS Millions of instructions per second. clock speed A series of electronic pulses produced at a predetermined rate that affects machine cycle time. microcode Predefined, elementary circuits and logical operations that the processor performs when it executes an instruction. Processing Characteristics and Functions Because having efficient processing and timely output is important, organizations use a variety of measures to gauge processing speed. These measures include the time it takes to complete a machine cycle and clock speed. Machine Cycle Time As we ve seen, the execution of an instruction takes place during a machine cycle. The time in which a machine cycle occurs is measured in fractions of a second. Machine cycle times are measured in microseconds (one-millionth of one second) for slower computers to nanoseconds (one-billionth of one second) and picoseconds (one-trillionth of one second) for faster ones. Machine cycle time also can be measured in terms of how many instructions are executed in a second. This measure, called MIPS, stands for millions of instructions per second. MIPS is another measure of speed for computer systems of all sizes. Clock Speed Each CPU produces a series of electronic pulses at a predetermined rate, called the clock speed, which affects machine cycle time. The control unit portion of the CPU controls the various stages of the machine cycle by following predetermined internal instructions, known as microcode. You can think of microcode as predefined, elementary circuits and logical operations that the processor performs when it executes an instruction. The control unit executes the microcode in accordance with the electronic cycle, or pulses of the CPU clock. Each microcode instruction takes at least the same amount of time as the interval between pulses. The shorter the interval between pulses, the faster each microcode instruction can be executed (see Figure 3.3). Figure 3.3 Clock Speed and the Execution of Microcode Instructions A faster clock speed means that more microcode instructions can be executed in a given time period. Amplitude 1 cycle 3 cycles Time (in seconds) hertz One cycle or pulse per second. megahertz (MHz) Millions of cycles per second. gigahertz (GHz) Billions of cycles per second. bit Binary digit 0 or 1. Clock speed is often measured in megahertz. As seen in Figure 3.3, a hertz is one cycle or pulse per second. Megahertz (MHz) is the measurement of cycles in millions of cycles per second, and gigahertz (GHz) stands for billions of cycles per second. The clock speed for personal computers can range from 200 MHz for computers bought in the mid-1990s to well over 3.2 GHz for the most advanced systems. 6 Because the number of microcode instructions needed to execute a single program instruction such as performing a calculation or printing results can vary, there is no direct relationship between clock speed measured in megahertz and processing speed measures such as MIPS and milliseconds. Although widely touted, clock speed is only meaningful when making speed comparisons between computer chips in the same family from the same manufacturer; comparing one Intel Pentium 4 chip with another, for example. Wordlength and Bus Line Width Data is moved within a computer system in units called bits. A bit is a binary digit 0 or 1. Another factor affecting overall system performance is the number of bits the CPU can

9 Hardware: Input, Processing, and Output Devices Chapter 3 95 process at one time, or the wordlength of the CPU. Early computers were built with CPUs that had a wordlength of 4 bits, meaning that the CPU was capable of processing 4 bits at one time. The 4 bits could be used to represent actual data, an instruction to be processed, or the address of data to be accessed. The 4-bit limitation was quite confining and greatly constrained the power of the computer. Over time, CPUs have evolved to 8-, 16-, 32-, and 64-bit machines with dramatic increases in power and capability. Computers with larger wordlengths can transfer more data between devices in the same machine cycle. They can also use the larger number of bits to address more memory locations and hence are a requirement for systems with certain large memory requirements. A 64-bit machine allows the CPU to directly address 18 quintrillion (a billion billion) unique address locations compared with 4.3 billion for a 32-bit processor. The ability to directly access a larger address space is critical for multimedia, imaging, and database applications; however, the computer s operating system and related application software must also support 64-bit technology to achieve the full benefit of the 64-bit architecture. Data is transferred from the CPU to other system components via bus lines, the physical wiring that connects the computer system components. The number of bits a bus line can transfer at any one time is known as bus line width. Bus line width should be matched with CPU wordlength for optimal system performance. It would be of little value, for example, to install a new 64-bit bus line if the system s CPU had a wordlength of only 16. Assuming compatible wordlengths and bus widths, the larger the wordlength, the more powerful the computer. Because all these factors machine cycle time, clock speed, wordlength, and bus line width affect the processing speed of the CPU, comparing the speed of two different processors even from the same manufacturer can be confusing. Although the megahertz rating has important consequences for the design of a computer system and is therefore important to the computer design engineer, it is not necessarily a good measure of processor performance, especially when comparing one family of processors with the next or when making comparisons between manufacturers. Chip makers such as Intel, Advanced Micro Devices, and Sun Microsystems have developed a number of benchmarks for speed. To ensure objective comparisons, many people prefer to use general computer system benchmarks such as SYSmark, distributed by a consortium called the Business Application Performance Corporation, whose members include hardware and software manufacturers such as AMD, Dell, Hewlett-Packard Co., IBM, Intel, and Microsoft, and industry publications such as Computer Shopper and ZDNet. For less-technical measures of performance, popular computer journals (such as PC Magazine and PC World) often rate personal computers on price, performance, reliability, service, and other factors. Physical Characteristics of the CPU CPU speed is also limited by physical constraints. Most CPUs are collections of digital circuits imprinted on silicon wafers, or chips, each no bigger than the tip of a pencil eraser. To turn a digital circuit within the CPU on or off, electrical current must flow through a medium (usually silicon) from point A to point B. The speed at which it travels between points can be increased by either reducing the distance between the points or reducing the resistance of the medium to the electrical current. Reducing the distance between points has resulted in ever-smaller chips, with the circuits packed closer together. In the 1960s, shortly after patenting the integrated circuit, Gordon Moore, former chairman of the board of Intel (the largest microprocessor chip maker), formulated what is now known as Moore s Law. This hypothesis states that transistor (the microscopic on/off switches, or the microprocessor s brain cells) densities on a single chip will double every 18 months. Moore s Law has held up amazingly for nearly four decades. In 2003, Moore himself forecast that the steady growth in the density and performance of microprocessors may go on only for the next 8 to 12 years. A key problem will be the need to narrow the minimum width of basic circuit features of a chip, which today are 90 nanometers (one billionth of a meter, 10-9 meter) wide in the most advanced chip manufacturing process. 7 (The 90-nanometer designation refers to the width of the smallest circuit lines on the chip. The actual features on 90-nanometer chips can be quite small, down to around 45 nanometers for some of the smallest structures on the chip.) 8 Chip makers then will have to wordlength The number of bits the CPU can process at any one time. bus line The physical wiring that connects the computer system components. Moore s Law A hypothesis that states that transistor densities on a single chip will double every 18 months.

10 96 Part 2 Information Technology Concepts Figure 3.4 Moore s Law (Source: Data from Moore s Law: Overview, Intel Web site at accessed February 11, 2004; and Gary Anthes, Microprocessors March On, Computerworld, March 10, 2003, 1,000,000,000 find a new technology to replace the semiconductor if they are to keep up this rate of improvement. As silicon transistors grow smaller there will be a billion on a single chip by 2008 power dissipation becomes nearly impossible to control, and even cosmic rays can cause random processing errors. 9 According to Nathan Brockwood, principal analyst at research company Insight 64, Every generation [of new chips] requires greater investment in [research and development] and manufacturing to make it work, because the low-hanging fruit in terms of semiconductor production was harvested years ago. 10 In addition to increased processing speeds, Moore s Law has had an impact on costs and overall system performance. As seen in Figure 3.4, the number of transistors on a chip continues to climb. NUMBER OF TRANSISTORS 100,000,000 10,000,000 1,000, ,000 10, Pentium Pentium II Pentium III Pentium , YEAR INTRODUCED superconductivity A property of certain metals that allows current to flow with minimal electrical resistance. optical processors Computer chips that use light waves instead of electrical current to represent bits. Researchers are taking many approaches to continue to improve the performance of computers. One approach is to substitute superconductive material for the silicon in computer chips. Superconductivity is a property of certain metals that allows current to flow with minimal electrical resistance. Traditional silicon chips create some electrical resistance that slows processing. Chips built from less-resistant superconductive metals offer increases in processing speed. Materials other than silicon, including carbon and gallium arsenide (GaAs), are used in the development of special-purpose chips. Several companies are experimenting with chips called optical processors, which use light waves instead of electrical current to represent bits. The primary advantage of optical processors is their speed. Lenslet, an Israeli-based start-up firm, has developed an optical computer that is capable of performing 8 trillion operations per second (teraflops). The company s prototype is fairly large and bulky, but the goal is to shrink it to a single chip by The powerful processor will be put to use in such applications as high-resolution radar, electronic warfare, luggage screening at airports, video compression, weather forecasting, and cellular phone base stations. 11 IBM and other companies are turning to strained silicon, a technique that boosts performance and lowers power consumption by stretching silicon molecules farther apart, thus allowing electrons to experience less resistance and flow up to 70 percent faster, which can lead to chips that are up to 35 percent faster without having to shrink the size of transistors. 12 Another advancement is the development of the carbon nanotube, so called because it is a pure carbon tube made of hexagonal structures 1 to 3 nanometers in diameter. The

11 Hardware: Input, Processing, and Output Devices Chapter 3 97 PMC-Sierra s RM9000 family of MIPS-RISC processors provides high-performance and low-power solutions for embedded applications such as networking, printing, workstation, and consumer devices. (Source: Courtesy of PMC-Sierra, Inc.) nanotubes can be used to form the tiny circuits for computer components. The practical use of nanotubes may not occur until sometime in the 2010 decade because of manufacturing difficulties and other complications. Scientists at the Los Alamos National Laboratory have taken miniaturization to the extreme. They are experimenting with using radio waves to manipulate individual atoms into executing a simple computer program. Their goal is to be able to manipulate thousands of atoms and build a computer many times smaller yet more powerful than any computer currently in existence. Complex and Reduced Instruction Set Computing Processors for many personal computers are designed based on complex instruction set computing (CISC), which places as many microcode instructions into the central processor as possible. In the mid-1970s John Cocke of IBM recognized that most of the operations of a CPU involved only about 20 percent of the available microcode instructions. This led to an approach to chip design called reduced instruction set computing (RISC), which involves reducing the number of microcode instructions built into a chip to this essential set of common microcode instructions. RISC chips are faster than CISC chips for processing activities that predominantly use this core set of instructions because each operation requires fewer microcode steps prior to execution. Most RISC chips use pipelining, which, as mentioned earlier, allows the processor to execute multiple instructions in a single machine cycle. With less sophisticated microcode instruction sets, RISC chips are also less expensive to produce and are quite reliable. Gates Corporation, which makes belts, hoses, and automotive products, moved 42 business applications off its aging mainframe computers onto two RISC-based HP Superdome servers running an Oracle ERP program and HP systems-management software. Gates has 64 processors installed in each Superdome, which can be expanded to 256 processors. The move was made to boost processing capacity and system reliability. 13 In June 2003, Apple introduced the PowerPC G5, the world s first 64-bit RISC processor for desktop computers. By almost any benchmark, RISC processors run faster than Intel s Pentium processor. And because RISC chips have a simpler design and require less silicon, they are cheaper to produce. The PowerPC chip is designed to provide portable and desktop personal computers the processing power normally associated with much more expensive computers. For example, the Macintosh PowerPC G5 has the ability to make functions such as voice recognition, dictation, pen input, and touchscreens practical. The PowerPC G5 can come with two IBM PowerPC 970 microprocessors with a 64-bit architecture. Sun Microsystems Sparc chip is another example of a RISC processor. When selecting a CPU, organizations must balance the benefits of speed with cost. CPUs with faster clock speeds and machine cycle times are usually more expensive than slower ones. This expense, however, is a necessary part of the overall computer system cost, for the CPU complex instruction set computing (CISC) A computer chip design that places as many microcode instructions into the central processor as possible. reduced instruction set computing (RISC) A computer chip design based on reducing the number of microcode instructions built into a chip to an essential set of common microcode instructions.

12 98 Part 2 Information Technology Concepts is typically the single largest determinant of the price of many computer systems. CPU speed can also be related to complexity. Having a less complex code, as in the case of RISC chips, not only can increase speed and reliability but can also reduce chip manufacturing costs. byte (B) Eight bits that together represent a single character of data. Memory Characteristics and Functions Main memory is located physically close to the CPU, but not on the CPU chip itself. It provides the CPU with a working storage area for program instructions and data. The chief feature of memory is that it rapidly provides the data and instructions to the CPU. Storage Capacity Like the CPU, memory devices contain thousands of circuits imprinted on a silicon chip. Each circuit is either conducting electrical current (on) or not (off). Data is stored in memory as a combination of on or off circuit states. Usually 8 bits are used to represent a character, such as the letter A. Eight bits together form a byte (B). Following is a list of storage capacity measurements. In most cases, storage capacity is measured in bytes, with one byte usually equal to one character. The contents of the Library of Congress, with over 126 million items and 530 miles of bookshelves, would require about 20 petabytes of digital storage. Name Abbreviation Number of Bytes Byte B 1 Kilobyte KB 2 10 or approximately 1,024 bytes Megabyte MB 2 20 or 1,024 kilobytes (about 1 million) Gigabyte GB 2 30 or 1,024 megabytes (about 1 billion) Terabyte TB 2 40 or 1,024 gigabytes (about 1 trillion) Petabyte PB 2 50 or 1,024 terabytes (about 1 quadrillion) Exabyte EB 2 60 or 1,024 petabytes (about 1 billion billion, or 1 quintillion) random access memory (RAM) A form of memory in which instructions or data can be temporarily stored. Types of Memory There are several forms of memory, as shown in Figure 3.5. Instructions or data can be temporarily stored in random access memory (RAM). RAM is temporary and volatile RAM chips lose their contents if the current is turned off or disrupted (as in a power surge, brownout, or electrical noise generated by lightning or nearby machines). RAM chips are mounted directly on the computer s main circuit board or in other chips mounted on peripheral cards that plug into the computer s main circuit board. These RAM chips consist of millions of switches that are sensitive to changes in electric current. Figure 3.5 Basic Types of Memory Chips Memory Types Volatile Nonvolatile RAM ROM SDRAM DRAM EDO PROM EPROM RAM comes in many different varieties. One version is extended data out, or EDO RAM, which is faster than older types of RAM memory. Another kind of RAM memory is called dynamic RAM (DRAM) and is based on single-transistor memory cells. SDRAM, or synchronous DRAM employs a minimum of four transistors per memory cell and needs high or low voltages at regular intervals every two milliseconds (two one-thousands of a second) to retain its information. Compared with EDO RAM, SDRAM provides a faster transfer speed between the microprocessor and the memory.

13 Hardware: Input, Processing, and Output Devices Chapter 3 99 Another type of memory, ROM, an acronym for read-only memory, is usually nonvolatile. In ROM, the combination of circuit states is fixed, and therefore its contents are not lost if the power is removed. ROM provides permanent storage for data and instructions that do not change, such as programs and data from the computer manufacturer, including the instructions that tell the computer how to start up when power is turned on. There are other types of nonvolatile memory as well. Programmable read-only memory (PROM) is a type in which the desired data and instructions and hence the desired circuit state combination must first be programmed into the memory chip. Thereafter, PROM behaves like ROM. PROM chips are used in situations in which the CPU s data and instructions do not change, but the application is so specialized or unique that custom manufacturing of a true ROM chip would be cost prohibitive. A common use of PROM chips is for storing the instructions to popular video games, such as those for GameBoy and Xbox. Game instructions are programmed onto the PROM chips by the game manufacturer. Instructions and data can be programmed onto a PROM chip only once. Erasable programmable read-only memory (EPROM) is similar to PROM except, as the name implies, the memory chip can be erased and reprogrammed. EPROMs are used when the CPU s data and instructions change, but only infrequently. An automobile manufacturer, for example, might use an industrial robot to perform repetitive operations on a certain car model. When the robot is performing its operations, the nonvolatility and rapid accessibility to program instructions offered by EPROM is an advantage. Once the model year is over, however, the EPROM controlling the robot s operation will need to be erased and reprogrammed to accommodate a different car model. Over the past decade, microprocessor speed has doubled every 18 months, but memory performance has not kept pace. In effect, memory has become the principal bottleneck to system performance. Thus, microprocessor manufacturers are working with memory vendors to create memory that can keep up with the performance of faster processors and bus architectures. One approach that has been taken is to employ cache memory, a type of high-speed memory that a processor can access more rapidly than main memory (see Figure 3.6). Frequently used data is stored in easily accessible cache memory instead of slower memory such as RAM. Because there is less data in cache memory, the CPU can access the desired data and instructions more quickly than if it were selecting from the larger set in main memory. The CPU can thus execute instructions faster, and the overall performance of the computer system is improved. There are three types of cache memory. The Level 1 (L1) cache is on the CPU chip. The Level 2 (L2) cache memory can be accessed by the CPU over a high-speed dedicated bus interface. The latest processors go a step further and place the L2 cache directly on the CPU chip itself and provide high-speed support for a tertiary Level 3 (L3) external cache. Deerfield, often called the Itanium 2 chip, is the low-power version of Intel s 64-bit servers and was introduced with 1.5 MB of Level 3 cache. 14 read-only memory (ROM) A nonvolatile form of memory. cache memory A type of high-speed memory that a processor can access more rapidly than main memory. Figure 3.6 CPU Miss Cache Controller Hit Cache Memory Typically 4 MB Typically 256 KB or More Memory (Main Store) Cache Memory Processors can access this type of high-speed memory faster than main memory. Located on or near the CPU chip, cache memory works in conjunction with main memory. A cache controller determines how often the data is used and transfers frequently used data to cache memory, then deletes the data when it goes out of use. When the processor needs to execute an instruction, it looks first in its own data registers. If the needed data is not there, it looks to the L1 cache, then to the L2 cache, then to the L3

14 100 Part 2 Information Technology Concepts cache. If the data is not in any cache, the CPU requests the data from main memory. It might not even be there, in which case the system has to retrieve the data from secondary storage. It can take from one to three clock cycles to fetch information from the L1 cache, while the CPU waits and does nothing. It takes 6 to 12 cycles to get data from an L2 cache on the processor chip. It can take dozens of cycles to fetch data from an L3 cache and hundreds of cycles to fetch data from secondary storage. This hierarchical arrangement of memory helps bridge a widening gap between processor speeds, which are increasing at roughly 50 percent per year, and DRAM access rates, which are climbing at only 5 percent per year. Costs for memory capacity continue to decline. When considered on a megabyte-tomegabyte basis, memory is still considerably more expensive than most forms of secondary storage. Memory capacity can be important in the effective operation of a CBIS. The specific applications of a CBIS determine the amount of memory required for a computer system. For example, complex processing problems, such as computer-assisted product design, require more memory than simpler tasks such as word processing. Also, because computer systems have different types of memory, other programs may be needed to control how memory is accessed and used. In other cases, the computer system can be configured to maximize memory usage. Before additional memory is purchased, all these considerations should be addressed. multiprocessing The simultaneous execution of two or more instructions at the same time. coprocessor Part of the computer that speeds processing by executing specific types of instructions while the CPU works on another processing activity. massively parallel processing A form of multiprocessing that speeds processing by linking hundreds or thousands of processors to operate at the same time, or in parallel, with each processor having its own bus, memory, disks, copy of the operating system, and applications. Multiprocessing A number of forms of multiprocessing involve the simultaneous execution of two or more instructions at the same time. One form of multiprocessing involves coprocessors. A coprocessor speeds processing by executing specific types of instructions while the CPU works on another processing activity. Coprocessors can be internal or external to the CPU and may have different clock speeds than the CPU. Each type of coprocessor performs a specific function. For example, a math coprocessor chip can be used to speed mathematical calculations, and a graphics coprocessor chip decreases the time it takes to manipulate graphics. Massively Parallel Processing Another form of multiprocessing, called massively parallel processing, speeds processing by linking hundreds and even thousands of processors to operate at the same time, or in parallel. Each processor includes its own bus, memory, disks, copy of the operating system, and application software. With parallel processing, a business problem (such as designing a new product or piece of equipment) is divided into several parts. Each part is solved by a separate processor. The results from each processor are then assembled to get the final output (see Figure 3.7). Massively parallel processing systems can coordinate large amounts of data and access them with greater speed than was previously possible. The most frequent business uses for massive parallel processing include modeling, simulation, and the analysis of large amounts Figure 3.7 Massively Parallel Processing Massively parallel processing involves breaking a problem into various subproblems or parts, then processing each of these parts independently. The most difficult aspect of massively parallel processing is not the simultaneous processing of the subproblems but the logical structuring of the problem into independent parts. Processor A Part A Solution A Part B Processor B Solution B Processing Job Part C Processor C Solution C Part D Solution D Part E Processor D Solution E Processor E Final Results

15 Hardware: Input, Processing, and Output Devices Chapter of data. In today s challenging marketplace, consumers are demanding increased product features and a whole array of new services. These consumer demands have forced companies to find more effective and insightful ways of gathering and analyzing information, not just about existing customers but also potential customers. Collecting and organizing this enormous amount of data is difficult. Massively parallel processing can access and analyze the data to create the information necessary to build an effective marketing program that can give the company a competitive advantage. Ford Motor Company is interested in the use of massively parallel processing to predict driver and passenger injuries in accident scenarios. The computer could compute and predict the damage done to human organs. Today s analyses with test dummies are very crude. They can determine at a basic level whether a certain type of crash is survivable. But occupant injury analysis takes much more computing power than is available now. 15 Symmetrical multiprocessing (SMP) is another form of parallel processing in which multiple processors run a single copy of the operating system and share the memory and other resources of one computer. Sharing resources creates more overhead than a singleprocessor system or the massively parallel processing system. As a result, the processing capability of SMP systems isn t proportionally greater than that of single processor systems (i.e., the capability of an SMP processor with two processors is less than twice the speed of a single processor). SMP has been implemented in the Sun Microsystems UltraSparc and Sparcserver, IBM Alpha, Macintosh PowerPC, and Intel chips. Cendant Corporation operates Internet-based travel sites, including CheapTickets.com and Galileo, a corporate reservations site. It recently moved its airline fare system to IBM e- Server SMP systems to deliver fares to customers faster. In addition, the project saved the company tens of millions of dollars in hardware maintenance and programming costs. The Galileo 360 efares system now runs on more than 100 clustered IBM eserver x440 and x445 systems. These computers are high-performance four-way and eight-way SMP servers that use Intel Xeon processors, which Cendant links together in powerful clusters. With the new SMP systems, Cendant reduced the hours of preprocessing work necessary to post the new fares issued by airlines six times a day. End users can access airline updates immediately and get the new fares faster. 16 Grid Computing Grid computing is the use of a collection of computers, often owned by multiple individuals or organizations, to work in a coordinated manner to solve a common problem. Grid computing is one low-cost approach to massively parallel processing. The grid can include dozens, hundreds, or even thousands of computers that run collectively to solve extremely large parallel processing problems. Key to the success of grid computing is a central server that acts as the grid leader and traffic monitor. This controlling server divides the computing task into subtasks and assigns the work to computers on the grid that have (at least temporarily) surplus processing power. The central server also monitors the processing, and if a member of the grid fails to complete a subtask, it will restart or reassign the task. When all the subtasks are completed, the controlling server combines the results and advances to the next task until the whole job is completed. The types of computing problems most compatible with grid computing are those that can be divided into problem subsets that can run in parallel. Such problems are frequently encountered in scientific and engineering computing. Mission-critical or time-sensitive applications should not be considered for grid computing because security, reliability, and performance cannot be guaranteed currently. Business issues also arise. For example, if an application is to run on 25 computers owned by 15 different organizations, how do you define who pays the costs and how those costs are allocated among the organizations? In a grid experiment nicknamed the Big Mac, Virginia Tech students and staff linked 1,100 Macintosh G5 Power Mac computers to form the world s third-fastest supercomputer, capable of performing 10.3 trillion operations per second. The processors are linked by a high-speed network called Infiniband that allows them to break up major calculations and analyze each part at the same time. The entire system cost about $7 million. 17 symmetrical multiprocessing (SMP) Another form of parallel processing in which multiple processors run a single copy of the operating system and share the memory and other resources of one computer. grid computing The use of a collection of computers, often owned by multiple individuals or organizations, to work in a coordinated manner to solve a common problem.