Topic 1 - Introduction to Computer Hardware Architecture
IBM PC Institute ibm.com/pc/training
Introduction to Computer Hardware Architecture
TXW02-R15
July 2001
1
Topic 1 - Introduction to Computer Hardware Architecture
Objectives: Computer Hardware Architecture Upon completion of this section, you will be able to: 1. Identify the types of computers and their key differentiating features 2. List key features of information appliances 3. Describe the concept of a thin client 4. Identify different ways that peer-to-peer computing is implemented 5. Define various PC architecture terminology including computer layers, controllers, and buses in a computer 6. Identify common industry standards and the objective of benchmarks used with computer systems
TXW02-R15
July 2001
2
Topic 1 - Introduction to Computer Hardware Architecture
Types and Features of Computers
Notebook - Mobile - Easy access to data Designs - Ultraportable - Full function
IBM Brands - ThinkPad - ThinkPad i Series - WorkPad (PC Companion)
Desktop
Server
- Non-mobile - Constant view of data
- High security - Process and file data
Designs - Full function - Legacy-free - All-in-one - Thin client IBM Brands - IBM PC 300 - NetVista - IntelliStation - Network Station
Designs - Tower - Rack - 1U rack - Blades IBM Brands - Netfinity - IBM ~ xSeries
Types and Features of Computers The three main types of computers (or PCs) are notebooks, desktops, and servers. Notebooks are optimized for traveling and for mobile users who need easy access to data. Desktops are for users who work in one place and who need access to data on the desktop or through a network. Servers are systems that process data (application servers), store data (file servers), or serve clients in various ways. Servers are in secure areas because so many users are dependent on their function. Each computer has different implementations or designs. Notebooks can be categorized as ultraportable (around 3 pounds) or full function, which IBM implements with the ThinkPad and ThinkPad i Series. PC companions are another type of mobile device, which includes the IBM WorkPad. Desktop systems designs have traditionally been full-function systems, but legacy-free and allin-one designs are available. IBM PC 300 and IntelliStation systems are full function systems. IBM NetVista includes full function, legacy free, and all-in-one. While any notebook or desktop system could be classified as a thin client, the IBM Network Station is exclusively a thin client. Server designs include a tower (meaning it can stand on a floor), rack-based (meaning it must be installed in a rack), or 1U rack-based (meaning it is a thin 1.75" (1U) high server that must fit in a rack). IBM used the Netfinity brand for Intel-based server systems, then migrated to the ~ xSeries in October 2000.
TXW02-R15
July 2001
3
Topic 1 - Introduction to Computer Hardware Architecture
PC Workstations versus UNIX Workstations Desktop systems with high-end subsystems for graphics and engineering users are often referred to as workstations. They typically run a UNIX-based operating system instead of a Microsoft Windows-based operating system. Below is a comparison of PC workstations (such as the IBM IntelliStation) and UNIX workstations. PC Workstations
UNIX Workstations
Basics
Minimum configuration of latest Intel processor; price starting around $2,000
Minimum configuration of Sun UltraSPARC or IBM PowerPC processor; more memory; higher-end graphics; price starting under $5,000
Strengths
Low price; familiar Windows interface and Windows applications run natively; ease of manageability
Weaknesses
Still lags behind UNIX counterparts in 3D graphics capabilities; large applications may overwhelm system performance; lack of third-party graphics applications
Strong graphics; large number of third-party applications available for engineering and financial modeling; processor power to spare Difficult to manage; price prohibitive in some cases; Microsoft Office applications don't run natively
TXW02-R15
July 2001
4
Topic 1 - Introduction to Computer Hardware Architecture
Types and Features of Computers: Differentiating Computer Features
Notebook - Size and weight - Power mgmt/battery - Screen type and size - PC Card slots - Docking station - Modular bay - Screen size
Desktop - Fastest uniprocessor - Graphics performance - Systems management - Graphics controller - Removable storage (CD-RW, DVD-RAM) - Chip set
Server - Many slots - Disk capacity/performance - Multiple processors (SMP) - Large memory capacity - Reliability - Data integrity (ECC memory) - Light Path Diagnostics - Hot swap, redundant fans and power supplies
Differentiating Computer Features Each type of computer has important characteristics that distinguish it from the other classes. Differentiating features of notebooks include the following: • Size and weight • Power management and battery life • Screen type and size • PC Card slots • Docking station • Sales presentation capability • Docking station and/or port replicator support • Integrated infrared • Number of bays that are within the unit • Modular bays Differentiating features of desktops: • Fastest uniprocessor • Graphics performance • Systems management • Graphics controller
TXW02-R15
July 2001
5
Topic 1 - Introduction to Computer Hardware Architecture
• Removable storage (CD-RW, DVD-RAM) • Chip set • 3D graphics boards Differentiating features of servers include the following: • Clustering capability • High-availability features • Management software • Network operating system tuning • Network operating system certification • Light Path Diagnostics • Hot-swap, redundant fans • Hot-swap, redundant power supplies • Server setup tool (ServerGuide) • Keyboard/video/mouse switch support • Cable Chaining Technology (C2T) • Chassis design for easy servicing and upgrading • RAID • Redundant power supply and fans • Fibre Channel support • 1U rack models
Server Setup Tool (ServerGuide) ServerGuide is a server setup tool for simplifying the installation and administration of IBM servers. ServerGuide contains large amounts of reference data that is invaluable to the administrator when performing such tasks as tuning the server after installation. ServerGuide also holds useful utilities, device drivers, and application software. The Setup and Installation CD is a bootable CD that allows customers to run hardware configuration and set up programs without creating diskettes. Based on detected hardware, a sequence of tasks are presented as a wizard. For the advanced user, an individual selection menu is available.
ServerGuide Binder
TXW02-R15
July 2001
6
Topic 1 - Introduction to Computer Hardware Architecture
Hot-Swap Disks Most xSeries servers have hot-swap disks. Hot-swappable disks in a RAID array can be changed on the fly, without having to shut down the server.
Hot-Swap Disk
Hot-Swap Fans Many xSeries servers have hot-swap fans. Some are redundant, so a single fan failure will still properly cool the system.
Hot-Swap Fan
TXW02-R15
July 2001
7
Topic 1 - Introduction to Computer Hardware Architecture
Light Path Diagnostics IBM's unique Light Path Diagnostics feature for xSeries servers sets the standard for Intel processor-based server maintenance and repair. When the component Predictive Failure Analysis (PFA) capability indicates potential problems, it alerts the system manager and switches on the indicator on the Light Path Diagnostics panel of the server. Most xSeries and Netfinity servers include LED indicators, both on the front panel and within the server, to quickly guide customers to potential problems. These tools work together to simplify and speed up the repair of failing or failed components. For example, if a memory DIMM fails, LEDs on the system board can help to quickly and easily locate an individual memory DIMM. System administrators and service personnel can quickly and easily identify failing components, potentially without even running diagnostics. Symptoms of pending failure can be subtle or intermittent, requiring that technicians painstakingly test to identify the specific device that is failing. Light Path Diagnostics simplify server maintenance by eliminating this time-consuming step.
Light Path Diagnostics on xSeries Server
Hot-Swap Power Supply Many xSeries servers have hot-swap power supplies. Some servers have redundant power supplies, so the server operates properly even with a failed power supply.
Hot-swap and Redundant Power Supplies in xSeries Server
TXW02-R15
July 2001
8
Topic 1 - Introduction to Computer Hardware Architecture
Management Software IBM includes IBM Director management software. This utility lets you inventory, monitor, and update the server from a remote console, and can monitor a server’s performance over long periods of time, telling you where bottlenecks are or when they may possibly arise. This software can monitor the Windows Event Log for certain software failures and gain full remote control of the system. It can manage RAID arrays, Fibre Channel connections, and server clusters. Since most IBM servers have an onboard Advanced System Management processor, IBM Director management software lets you manage a server when that server has been shut down or the OS is hung.
KVM Switch A KVM switch is an active device that allows multiple systems to share one keyboard, video display, and mouse. These switches save users money and desktop space, because multiple components are not necessary to run the systems. A KVM switch is typically used for rackbased servers.
WITHOUT KVM SWITCH More equipment to buy and more space to take up
4 Port KVM Switch
WITH KVM SWITCH Less equipment to buy and less space to take up
TXW02-R15
July 2001
9
Topic 1 - Introduction to Computer Hardware Architecture
Cable Chaining Technology (C2T) The IBM xSeries 330 and some xSeries 130 and 135 systems have a great new IBM technology. The unique, new Cable Chaining Technology (C2T) connects multiple rack-based servers, thereby saving time and expense by eliminating more than 120 cables from a full rack of systems.
C2T in Multiple xSeries 330 systems Keyboard Video Mouse
One C2T Interconnect Cable between systems (ships standard) One Cable Chaining Technology (C2T) Console or Breakout Cable (std or optional)
Each xSeries 330 in a rack is connected with an Interconnect Cable that is daisy chained between each of up to 42 systems in a rack. This cable ships standard with each system.
xSeries 330 Interconnect Cable
Cable Chaining Technology (C2T) Console Cable
For every configuration (whether one xSeries 330 or a full rack of systems), you will need to have one Cable Chaining Technology (C2T) Breakout or Console Cable that comes off the system to either a keyboard, video (monitor), and mouse or to a KVM switch. This Cable Chaining Technology (C2T) Console Cable may have to be purchased separately with some systems.
TXW02-R15
July 2001
10
Topic 1 - Introduction to Computer Hardware Architecture
1U rack models Servers for racks vary in height by a U measurement (a U is 1.75-inch height). Servers that are 1U from IBM and other vendors are popular for web sites because for web pages it is better to spread the load across multiple servers (horizontal scalability) than to increase the processing power of a centralized server (vertical scalability).
TXW02-R15
July 2001
11
Topic 1 - Introduction to Computer Hardware Architecture
Types and Features of Computers: Legacy-Free System • System without older, complicated devices - No serial, parallel, keyboard, mouse ports; no diskette
• Universal Serial Bus (USB) ports for plug and play devices • Easy to use; low cost; space saving • Initially desktop; notebooks to follow • IBM NetVista S Series (S40, S40p)
NetVista S40 with Options
NetVista S40
Legacy-Free System Legacy-free systems are designed as easy-to-use, low-cost, space-saving systems optimized for Windows 2000 or Millennium environments. The term legacy-free describes a system without older, complicated devices such as serial ports, parallel ports, keyboard ports, and mouse ports. There is no diskette drive; a customer could use an external USB diskette if necessary. All ports on the legacy-free system are plug and play Universal Serial Bus (USB) ports that can be used for all standard devices. IBM introduced the NetVista S40 and S40p in April 2000 as legacy-free desktops, which are full-featured desktops in a space saving design that is only 12 inches high, 4 inches wide, and 15 inches deep. These desktops are expandable with two low profile PCI slots and five USB ports. With no serial, parallel, keyboard, or mouse ports, these systems use a USB mouse and USB keyboard. Further, these systems have slide-out external bays, one screw for cover removal, and an optional cradle for holding an external bay or an optional WorkPad PC Companion.
NetVista S40 - Rear View with three USB ports (2 USB in Front)
TXW02-R15
July 2001
12
Topic 1 - Introduction to Computer Hardware Architecture
Types and Features of Computers: All-in-One Desktop • System unit and a flat panel monitor integrated into one unit • Space-saving design • Ports are plug-and-play Universal Serial Bus (USB) • IBM NetVista X Series (X40, X40i)
NetVista X40 side view
NetVista X40
All-in-One Desktop The All-in-One is a full-powered system and flat-panel monitor engineered into one desktop that offers a space saving design along with expansion capabilities. All ports on the All-in-One are plug-and-play Universal Serial Bus (USB) ports that can be used for all standard devices. IBM introduced the NetVista X40 and X40i in April 2000 as all-in-one systems, which are fullfeatured desktops that are only 16 inches high, 16 inches wide, and 10 inches deep. These systems include an integrated 15 inch TFT flat-panel monitor, which provides a viewing area equivalent to that of a 17 inch CRT. The X40 and X40i have legacy-free features with five USB ports and no serial or parallel ports (they do not have a keyboard or a mouse port). The systems have two low-profile PCI slots and a unique design that allows the CD-ROM and diskette drive to hide away. An optional radial arm allows the unit to mount on a cubicle wall or desk.
NetVista X40 - Rear View
TXW02-R15
NetVista X40 - CD-ROM/Diskette
July 2001
13
Topic 1 - Introduction to Computer Hardware Architecture
Key advantages of the All-in-One system includes its compact size and a power supply that does not require much power. It also discourages configuration changes since it may have few, if any, slots or bays. The disadvantages include lack of expansion bays and multiple slots. The price range tends to be slightly higher than a separate system unit and monitor. Although the term All-in-One implies only one unit, the keyboard is always a separate component.
TXW02-R15
July 2001
14
Topic 1 - Introduction to Computer Hardware Architecture
Types and Features of Computers: File Server versus Application Server File server
Application server
• Remote disks that service requests for files • Transfers files to client • Appears on client as a logical drive (e.g., F:) • Normally on a LAN
• Processes a request and returns only answer (not a file) • Typically processes database request • Client sees application front end only • On LAN or Web
File Server versus Application Server A file server is a server that sends files to a client. A client will typically not need to know the file is on a different computer as it will appear within the file system on their system. For example, the user will see a logical drive such as F: and all the files on the F: drive. When the user selects a file, the file server gets the file and transfers it across the network into the memory of the client system. An application server is a server that runs an application on the server due to a client request and returns only a small amount of data to the client. Application servers are often part of Web sites that generate personalized or dynamic content. A user will enter data into a browser, which is used by the Web application server to run business logic on the input (e.g., looking up customer information in a database). The application server will output HTML and send it back through the Web server to the end user. Originally, Web applications were typically designed with custom programs written in C and using the Web server’s Common Gateway Interface (CGI) API. CGI is still common on the Web. Because CGI requires spawning a new instance of a program each time it processes a request however, as well as in the interest of better security, CGI’s prevalence is giving way to two other types of application servers—page-based and component-based.
TXW02-R15
July 2001
15
Topic 1 - Introduction to Computer Hardware Architecture
Page-based Web application servers tend to be used on sites that need interactivity but don’t have large budgets or programmer resources. Although it is easier to develop for and deploy dynamic sites with page-based scripting, these systems combine the code used to display the page (the presentation logic) with the code that performs the business logic (which usually gets or sets the appropriate database information). Examples of products for page-based application servers are Macromedia ColdFusion, Apache Software Foundation JSP (Java Server Pages) and Tomcat, and Zend Technologies Zend and PHP, and Microsoft Active Server Pages (ASP). Higher-end application servers use a component-based development model, which separates business logic from presentation logic. This compartmentalization lets programmers retool large Web sites quickly as new code modules can be dropped in as needed, which means avoiding rebuilding an entire system or searching through the scripts to find what has to be updated. Because of that, these products are the right choices for complex, often team-written Web applications such as high-volume e-commerce sites. Examples of products include IBM WebSphere, iPlanet Application Server, Microsoft IIS with COM+ and ASP, and BEA WebLogic.
File server: performs as remote disks that service requests for files from clients and transfer entire files to the client across the network.
Application server: sends only the results to clients. Network traffic is reduced because these servers do the processing.
User requests the application to open a file stored on F drive.
User requests a list of 3000 records sorted by part number.
The client's networking software redirects the request to the file server's disk.
The client sends a request in SQL format to the application server running a database management back end.
The file server gets the file and transfers it across the network into the memory of the client PC.
The application server does searches, sorts, calculations, and indexing resulting from SQL requests and transfers the results to the client.
Client PC running spreadsheet
TXW02-R15
Client PC running database front end
July 2001
16
Topic 1 - Introduction to Computer Hardware Architecture
Information Appliances • Specialized servers dedicated to one task • Extremely easy to install and add to infrastructure • Use specialized and proprietary components • Embedded OS with no administration • Services and applications are preinstalled and preconfigured • Use web browser for configuration and administration
IBM Whistle InterJet
Information Appliances Information Appliances (also called network appliances or thin servers) are a growing new class of dedicated devices that include Internet servers, proxy/caching servers, database servers, mail servers, and file servers. Since they are easy and quick to install, they are popular for small businesses. They provide a big cost saving over time in terms of management and administration. All configuration is done through a Web browser as an embedded operating system and is transparent to users. They are typically sealed systems requiring little IT expertise and minimal user intervention. They provide support for multiple versions of every leading operating system (not tied to a specific OS). Many are upgradeable via FTP or downloading code from the Internet. They should not require any client software. Information appliances contrast with traditional multipurpose servers. One example is the multipurpose server appliance that provides Internet connectivity, firewall servers, web access, and e-mail in a single, bundled solution. Some offer virtual private network (VPN) connections, print- and file-server functions, and the ability to operate as a web server. The product usually consists of an embedded OS with features like e-mail and Internet access managed through a browser, network-connected hardware, a disk, a processor, routing software, and a modem. Most appliances use a version of Linux running on an Intel processor. Those who purchase appliances place less emphasis on raw performance than on available features such as ease of use, flexible management, and reliability. For example, IBM acquired Whistle Communications, Inc., in 1999 and is reselling the Whistle InterJet, which is a thinserver appliance that sits on a LAN. The device provides a firewall and can host up to 100 IP addresses. It is sold to small businesses for messaging and Web access for a monthly service fee. The InterJet runs FreeBSD (an open-source version of Unix that predates Linux).
TXW02-R15
July 2001
17
Topic 1 - Introduction to Computer Hardware Architecture
Information Appliances: Examples Examples • Internet server appliance (Whistle InterJet) • Cache appliance • Enterprise database appliance • Enterprise mail appliance • Firewalls • Multipurpose server appliance • Transaction acceleration appliance
IBM 5194 TotalStorage Network Attached Storage 200
IBM Products • IBM xSeries 130 and 135 (Appliance Server) • IBM 5194 TotalStorage Network Attached Storage 200
Information Appliances: Examples Information appliances or servers come in many varieties. Web hosting servers combine software and hardware needed to host multiple Web sites. Web and media caching servers accelerate the delivery of Internet content to end users and save bandwidth by eliminating redundant network traffic on the WAN. Traffic management servers balance server traffic by making priority delivery decisions for optimal server farm and data center performance. VPN server appliances provide virtual private network service to connect employees, customers and e-business networks; serve as an alternative to long-distance dial-in, leased-line or frame relay connections. A cache appliance acts as an intermediary between requesting clients and servers on the Internet or intranet. Key web pages are stored on the cache appliance to be delivered to the end user as quickly as the network will allow. Software-based proxy servers can cache web pages also and are less expensive than a cache appliance but require hardware and software integration, are less reliable when run with other programs, lack fault tolerance, and need to be configured at each user's desktop. They therefore have higher support and maintenance costs. Other types of information appliances include enterprise database appliance, firewalls with or without software for setting up a virtual private network, enterprise mail appliance, transaction acceleration appliance, cluster administration appliance, e-commerce application hosting appliance, and network traffic and policy management appliance.
TXW02-R15
July 2001
18
Topic 1 - Introduction to Computer Hardware Architecture
Thin Clients • Any desktop or notebook can be configured as a thin client • Thin clients can reduce total cost of ownership . . . . . increase in function, performance, but also cost
Dumb Terminal S/390 terminal emulation
Thin Client
Full-function PC
Users: ● Task-based workers ● Customer service
Users: ● Content creators ● Knowledge workers
• Application and data reside and execute on server 2. Application processing on server
1. User input on client 3. Display info sent to client
Thin Clients Thin client systems are a class of systems that fall between dumb terminals and full-function PCs. Thin clients exist to reduce the total cost of ownership. In a thin client model, applications and data are stored on the server (usually called terminal server) which also processes the application and sends the graphical screen changes to the client. This model reduces upgrade and support costs while providing strong data integrity and security. The systems generally are small size and may not have slots or legacy connectors. Thin clients are easy to set up as initial configuration, software installation, software updates, data backup, and management can all be done remotely. Users cannot load software or devices that cause conflicts. All data resides on a secure server with a backup and redundancy plan. Thin clients are geared toward task-oriented workers such as reservation agents, health-care workers, administrative workers, or customer service personnel. They are not appropriate for knowledge workers such as professionals, engineers, developers, and graphic artists. Thin-client computing serves to reduce the management and maintenance costs of IT administration and, at the same time, enhance the quality and accessibility of applications to the end-user.
TXW02-R15
July 2001
19
Topic 1 - Introduction to Computer Hardware Architecture
The basic principle behind thin-client computing architecture is to separate the applicationprocessing layer from the user interface. On the server side, an application server handles the computational overhead and communicates to the client only the display information necessary for creating the interface, such as mouse positioning, window status, and text. The requirements of the client device do not need to be on par with the applications processing prerequisites. The device need only be capable of displaying a user interface. Thin-client technology extends the reach of application investments beyond the desktop to almost any type of device, including many non-Windows devices, wireless terminals, and information appliances, and allows applications to be run even in conditions of abject connectivity, because the amount of data transferred is dramatically reduced. End-users can thus conduct business more efficiently thanks to cross-platform availability of mission-critical resources, such as Unix-based business-to-business applications or business documents on the network. From the end-user's perspective, all of the applications, regardless of platform, are accessible from a singular environment on the thin-client device. With the burden of application processing pushed to a centralized server farm, thin-client processing extends the life cycle of existing infrastructure investments, reduces network traffic, and gives outdated systems access to state-of-the-art applications. Centralized, server-side application management reduces the total cost of ownership. Application updates can be rolled out more quickly and efficiently from a central point, and administrators maintain tighter control of clients.
TXW02-R15
July 2001
20
Topic 1 - Introduction to Computer Hardware Architecture
Thin Clients: Benefits and Issues Benefits Lower cost of support Clients can run Windows applications off server (even though do not support natively) High data security Increase performance by upgrading server only
Issues Configuration and licensing can be complex Fast network bandwidth is critical
Performance can be slow Need high-end server(s)
• Also called Windows-based Terminals • IBM Network Station is a thin client
IBM NetVista N2200
IBM NetVista N2800
IBM NetVista N2800
Thin Clients IBM offers the IBM NetVista Network Station systems as thin clients (such as the NetVista N2200 and N2800). The Network Station family of thin clients is designed to respond to a variety of business needs from accessing Windows applications on a server to running complete integrated business applications that incorporate the web, multimedia, Java, and Linux on the desktop. These systems use BSD Unix as the operating system and can run various applications in a variety of ways in conjunction with various server platforms. The systems ship with Network Station Manager (NSM) which is IBM's administration application which runs under NT, AIX, or OS/400. NSM allows you to specify how each unit will boot and which applications it can access. A Network Station can boot one of two ways. It can boot BSD Unix locally by taking the Kernal off an internal CompactFlash card. Or it can access a server loaded with NSM and pull a small OS Kernal over the network. Upon boot, it accesses an authentication file to verify the user name and password. A separate configuration file tells the server which applications to make available to the user. The configuration and authentication files can reside on the Flash Card or an NSM server. There are extensive applications available under BSD Unix. Tools can be run directly from the BSD desktop (including Java applications, browser, RealNetworks player). You can also connect to a remote host employing an ICA client for Windows applications, or you can run terminal emulation software. In October 2000, new models of the NetVista N2200 and N2800 were announced with TurboLinux software.
TXW02-R15
July 2001
21
Topic 1 - Introduction to Computer Hardware Architecture
Thin Clients: Method of Computing • User input interpreted and compressed, then sent (via RDP or ICA) to server • All application processing occurs at the server • Only changes to interface sent to client Terminal Server
Thin Client
Windows 2000 Terminal Services Windows CE, Windows 95, 98, Windows NT, 2000
Remote Desktop Protocol (RDP) (TCP/IP) Or
Or
Independent Computing Architecture (ICA) (TCP/IP, IPX, NetBIOS, SPX) Citrix MetaFrame
Same as above plus Unix, Mac, OS/2, DOS, ActiveX, Java
Thin Clients: Method of Computing Citrix and Microsoft have software for thin client models. This software is not an emulator but a client-neutral multi-user version of Windows NT/2000. Most 16-bit Windows applications, nearly all 32-bit Windows applications, and some DOS applications can run natively on the server at full speed. The server renders the text, bitmapped graphics, and RDP/ICA intercept calls to GDI and redirects them to the client. The server can store all of the user's files and configuration data so that he or she gets a personalized environment upon logon. Overlaid images and onscreen animations execute more slowly over the network to WTS client than they do natively. Microsoft Windows 2000 Terminal Services is bundled with Windows 2000 Server. This software only supports Windows clients (Windows CE, 95, 98, 2000). It uses Remote Desktop Protocol (RDP) technology to connect clients to terminal servers and to separate application logic (executed on the server) from the user interface (delivered to clients), so only keystrokes, mouse movements, and screen updates are sent across a network. It is limited to TCP/IP connections between client and server. There are no client fees for clients running Windows 2000 Professional, but there is a license fee for 9x and NT clients. Microsoft has two types of thin clients or Windows-based terminals (WBT): • WBT Standard: a Windows-based terminal based on the Windows CE operating system. In early 2000, IBM began shipping IBM Network Stations that were Standard Edition 1.5 compliant. • WBT Pro: Windows-based Terminal Professional, based on the Windows NT Embedded operating system
TXW02-R15
July 2001
22
Topic 1 - Introduction to Computer Hardware Architecture
Citrix MetaFrame and MetaFrame XP family of products runs on Windows NT or Windows 2000 Terminal Services (there is also a version that runs on some Unix versions). The previous version was Citrix WinFrame which worked with Windows NT 3.51. MetaFrame uses Independent Computing Architecture (ICA) technology to connect clients (including wireless clients) to terminal servers and to separate the application logic (executed on the server) from the user interface (delivered to clients). MetaFrame supports Windows clients and non-Windows clients (like Unix, Mac, OS/2, DOS, Java). The connection can be via IPX, NetBIOS, SPX, or TCP/IP. MetaFrame is basically a superset of Windows 2000 Terminal Services with support for more client types and protocol support. Citrix also sells Metaframe for Unix Operating Systems which provides access to Unix applications from non-Unix machines which eliminates both X Window System server and terminal emulation software. Both Microsoft and Citrix require a high-end server for acceptable performance. For example, the server should not provide other critical network services and multiple LAN adapter cards are recommended. NTFS is recommended for enhanced user session security. For each server, Microsoft suggests 30 users for heavy use to 75 users for light use. Besides the baseline memory requirement, Citrix recommends 4 MB to 8 MB for each additional user; Microsoft suggests 10 MB to 21 MB for each additional user. Management tools are included with the products. You must set up and administer clients, define session restrictions, set password and link encryption levels, install applications for multisession use, and set up user connections and permissions for applications. You must obtain per-user licenses or "number of connections per server" licenses for Windows 2000 Terminal Server users. MetaFrame requires a base license for a specified number of users, then you must obtain license packs for additional users beyond the base number. When setting up large environments, you will need multiple servers with copies of shared applications stored across many different servers.
TXW02-R15
July 2001
23
Topic 1 - Introduction to Computer Hardware Architecture
Peer-to-Peer Computing • Exchanging information among PCs without central servers • Gnutella - No central server; each PC is a servent - User searches contents of shared directories of peer members
• Napster - Central servers hold directories of files of registered users
• Groove - Each PC is a mini-server for workgroup for file sharing and collaboration
Many-to-many Web server E-mail server Corporate server
Peer-to-peer
Centralized
Peer-to-Peer Computing Peer-to-peer (P2P) is an alternative to the traditional centralized (or client-server) computing model. A centralized model utilizes server-based sharing and requires an intermediary such as a Web, e-mail, or corporate server. P2P has two or more computers linked for the purpose of sharing information files (locally or remotely) with each taking an equal role in the data-transfer process without the intervention of a central data. No server is needed to share information among systems; instead, each user's computer handles the serving, although those functions are hidden from the user. There are various P2P implementations. The two main implementations differ in the role of a central server. One model used by Gnutella does not use a central server. The other implementation uses central servers to hold directories and to direct traffic (not to store data), and it's used by Napster and Groove. Napster uses central servers to hold directories of music files stored on the PCs of registered users and to direct traffic among the users. When a Napster user requests a particular MP3 file, the central server displays a list of people who are currently connected who have the specified file. The requesting user can then link directly to any name on the list to download the file.
TXW02-R15
July 2001
24
Topic 1 - Introduction to Computer Hardware Architecture
One-to-One
Web server E-mail server Corporate server One-to-many
Centralized
Many-to-many
Peer-to-peer
Gnutella does not use a central server. Instead, each PC uses a piece of software called a servent to share files. Servent is a mix from the words server and client. On a Gnutella network, a PC notifies a connected PC of its presence. The notified PC then tells the computers it is connected to the first PC, and they in turn inform other members of the network in an expanding series of links. Once the members are aware of a PC, the user can search the contents of the shared directories of the peer network members. The search request follows the same path forward that was used when the computer announced its presence on the network, and the list of matching files travels that path back to the starting computer, which can then open a direct connection with a computer having the matching file and download the file. The Gnutella model can be used for any type of file, not just music. Groove turns every PC running the program into a mini-server and allows a workgroup, locally or widely distributed, to create virtual workspaces for sharing and real-time collaboration. Groove requires a downloaded client (similar to Napster). It uses true P2P connections but supports central relay servers when users are not connected and its real-time connections (instant messaging or conferencing) use direct connections between users.
TXW02-R15
July 2001
25
Topic 1 - Introduction to Computer Hardware Architecture
All data is stored on the peers' local drives, which act as mini-Web servers offering up shared spaces to team members. Changes to data or the interface are synchronized over P2P connections when users connect. A user can work off-line to perform tasks such as adding files, messages, and content, then sync such changes with other users on the next reconnect. Flypaper TeamSpace requires only a browser. Users can connect from any PC without downloading a client or plug-in. Uses pass-through P2P connections (like Napster service), sending even IM messages through Flypaper's central servers. Only direct file transfers skip the servers. All data is stored on the central servers, rather than locally. Shared TeamSpaces are completely hosted by Flypaper, as are individual configurations and personal settings. Synchronization is unnecessary; data is always current to everyone. Flypaper does not allow for off-line work. To see the projects that can be placed under the P2P umbrella, see the O'Reilly P2P Directory at www.oreillynet.com/pub/q/p2p_category. It shows projects ranging from file sharing to collaboration, messaging, resource distribution, and more. P2P technology is in an emerging state, and issues such as standards adherence, security, and manageability will need to be resolved before it becomes viable for the enterprise. P2P applications allow one user to access another's computer directly, sharing files and other resources is convenient, but hacking is convenient also. The Peer-to-Peer Working Group at www.peer-to-peerwg.org is a consortium for advancement of infrastructure best known practices for peer-to-peer computing. P2P is currently segmented by five different models listed in the table that follows. Five peer-to-peer models Atomistic
A true P2P architecture because it involves direct client-to-client connectivity with no server present because it has no server present; because it has no method of creating communications links based on data availability or user identity.
User-centered
User-centered applications utilize a directory to provide a way for users to make connections with other users on a network.
Data-centered
Data-centered applications allow users to search and access data held on other users' systems.
Web Mk 2
This is a convergence of the above three models with Web architectures and infrastructure. In this model, browsers evolve into user-configurable workspace managers that integrate these three types of P2P models. Multiple directory services can link users together on an ad hoc basis. Multiple indexes allow access to different forms of data whether it is on servers or clients.
Compute-centered
Instead of using a single large processor, an application's processing is divided among multiple clients and a server is used to coordinate the split processing. The distinction with this approach and parallel processing is that nodes are spread over the Internet and can be accessed on an as-needed basis.
Sun Microsystems has a P2P initiative called JXTA (pronounced "juxta") which provides an infrastructure for P2P applications. Its goal is interoperability so any system built or enabled with JXTA can communicate with one another.
TXW02-R15
July 2001
26
Topic 1 - Introduction to Computer Hardware Architecture
PC Architecture: Computer Layers User
• Layered structure allows for compatibility--bypassing layers increases performance • BIOS (basic input/output system) - located in flash memory (sometimes called EEPROM) - supports plug-and-play - supports power management
Applications API Operating System Device Driver Firmware
BIOS
Adapter
Hardware
• Extensible Firmware Interface (EFI) replaces BIOS for Itanium systems • Device driver - software to control a piece of hardware
EEPROM
Computer Layers Applications: the software with which a user typically interacts, such as those used for word processing, Web browsing, sending e-mail, and using spreadsheets. Operating system: a set of programs that provides an environment in which applications can run, allowing them easily to take advantage of the processor and I/O devices, such as disks or adapters. Basic input/output system (BIOS): a set of program instructions that activates system functions independently of hardware design (layer between the physical hardware and the operating system) and allows for software compatibility. BIOS is typically located in flash memory (EEPROM) on the systemboard. When a PC is started, the BIOS runs a power-on self-test (POST). It then tests the system and prepares the computer for operation by searching for other BIOSs on the plug-in boards and setting up pointers (interrupt vectors) in memory to access those routines. It then loads the operating system and passes control to it. The BIOS accepts requests from the drivers as well as the application programs. The BIOS supports plug-and-play and power management. BIOS vendors include IBM, Compaq, AMI, and Phoenix. Although there are several BIOS vendors, there are few differences among their products. Firmware is similar to BIOS. It is usually the layer of software that is between the device driver and adapter. It typically is on a EEPROM of an adapter card and can be upgraded with a diskette.
TXW02-R15
July 2001
27
Topic 1 - Introduction to Computer Hardware Architecture
Device driver: a type of software (which may be embedded in firmware) that controls or emulates devices attached to the computer such as a printer, scanner, diskette drive, hard disk, monitor, or mouse. Device drivers are typically loaded low into the memory of PCs at boot time. A device driver expands an operating system's ability to work with peripherals and controls the software routines that make peripherals work (a network card, a disk, printer). These routines may be part of another program (many applications include device drivers for printers), or they may be separate programs. Basic drivers come with the operating system, and drivers must be installed for each peripheral added. Windows Driver Model or Win32 Driver Model (WDM) by Microsoft allows a common driver for Windows 98, Windows NT, and Windows 2000. To preclude the problem of performing OS, BIOS, or driver updates before the OS or network drivers are loaded, a Preboot Execution Environment (PXE) allows the system to boot off the network. At boot, a PXE agent executes, and the PC gets an IP address from a DHCP server and then uses the BOOTP protocol to look for a PXE server. The PXE client is firmware implemented in BIOS (if LAN hardware is on the systemboard) or as a boot PROM (if LAN adapter). Programs, including those in the PXE environment, require system configuration and diagnostic information. A Systems Management BIOS (SMBIOS) is a chip that makes the necessary information available via BIOS calls that are available through the OS and in the preboot environment. Choosing the correct device driver for specific hardware is very important. Device drivers are also specific to an operating system. Some of the device drivers are supplied with the OS, and some are supplied with the hardware on CD. A technically-competent person should select a proper driver during installation. Selecting an incorrect device driver for a specific device (if it works at all) can cause poor performance or data loss. The hardware vendor normally provides the latest drivers on their Web site. Class driver: a driver that provides system-required, hardware-independent support for a given class of physical devices. Minidriver: a hardware-specific DLL that uses a Microsoft-provided class driver to accomplish most actions through functions calls and provides only device-specific controls. Under WDM, the minidriver uses the class driver's device object to make system calls. Minidrivers run at ring zero of the OS. This action speeds execution. Miniport driver: a device-specific kernel-mode driver linked to a Windows NT or WDM port driver, usually implemented as a DLL that provides an interface between the port driver and the system. Interrupt: the means by which external devices get processor attention. Exceptions are internal processor interrupts. The Intel architecture allows for 32 exceptions, including one nonmaskable external interrupt (NMI). There is an interrupt descriptor table that allows for up to 224 (usually 16) external maskable interrupts, multiplexed by a programmable interrupt controller to one INTR pin. Interrupt procedures either are part of the operating system or are installed in a device driver.
TXW02-R15
July 2001
28
Topic 1 - Introduction to Computer Hardware Architecture
Key to numbered steps: 1. The user presses on keyboard. 2. The code corresponding to pressed key is sent over the cable to the keyboard port on the systemboard.
Application Program
6.
3. The keyboard BIOS routine accepts the code, translates it into the letter n, and passes it to the operating system. 4. The operating system passes the keystroke to the application program and sends the n to the video BIOS (or the device driver). 5. The video BIOS sends the n to the graphics circuitry.
Operating System
4. 5.
7. Keyboard Cable
1.
6. The application program accepts the keystroke and instructs the operating system to look for the next keystroke.
BIOS
3.
Video Circuitry Keyboard Port Hardware
2.
Hardware
7. The n appears on the screen.
TXW02-R15
July 2001
29
Topic 1 - Introduction to Computer Hardware Architecture
Extensible Firmware Interface (EFI) This Extensible Firmware Interface (EFI) Specification describes an interface between the operating system (OS) and the platform firmware. The interface is in the form of data tables that contain platform-related information, and boot and runtime service calls that are available to the OS and its loader. Together, these provide a standard environment for booting an OS. The EFI specification is primarily intended for the next generation of IA-32 and IA-64 Intel architecture-based computers. Thus, the specification is applicable to a full range of hardware platforms from mobile systems to servers. EFI controls the computer’s boot configuration. The boot menu that once resided on a fragile disk is now located in durable flash memory (nonvolatile memory). The EFI has a shell mode that is similar to a DOS command prompt. From the prompt, you can perform basic file management tasks (including text editing), make configuration changes, or write scripts that execute at boot time. Unlike BIOS, the EFI can access DOS-formatted disk partitions, LS-120 SuperDisk disks, and standard CD-ROMs. Microsoft will support EFI as the only firmware interface to boot the 64-bit version of Windows for the Intel Architecture. Because the 64-bit version of Windows will not boot with BIOS or with System Abstraction Layer (SAL) alone, EFI is a requirement for all Intel Itanium-based systems to boot Windows.
Operating System
EFI OS Loader
(OTHER)
EFI Boot Services
SMBIOS
EFI Runtime Services
ACPI
Interfaces from other required specs
Platform Hardware EFI System Partition EFI OS LOADER
EFI Conceptual Overview
TXW02-R15
July 2001
30
Topic 1 - Introduction to Computer Hardware Architecture
The EFI brings support for a new partition structure called a Guided Partition Table (GPT), which uses a 128-bit unique Global Unique Identifier (GUID) as defined by the Wired for Management 2.0 specifications to the disk itself. The new structure is backward compatible with the master boot record structure and brings many advantages that can be exploited especially on large disks. A 64-bit logical block addressing mode is available, allowing for a potential maximum of 16 exabytes for the partition size to be defined inside a disk. With the new structure there is theoretically no limitation to the number of definable partitions. Redundancy is created by placing a copy of the original partition table at the very end of the disk, allowing for the recovery of the contents of the drive in case of physical failure of the initial blocks. LBA0
LBA1
First Usable Block
End Partition
LBAn
n
Partition Table Header
MBR
Partition Table Header
0 1 ..
Start Partition
Partition 1 0 1 .. n Start Partition
End Partition
Backup Partition Table
Primary Partition Table
TXW02-R15
July 2001
31
Topic 1 - Introduction to Computer Hardware Architecture
PC Architecture: Subsystems Major internal subsystems of a PC: • Processor (Pentium III) • L2 cache (256 KB) • Memory (128 MB) • Bus(es) (PCI, AGP) • Graphics controller (SVGA) • Disk controller (EIDE) and disk (40 GB) • Slots
Processor + L1 Cache
L2 Cache
MCH or GMCH Host Bridge
Memory
AGP Slot for SVGA Graphics PCI Slots
Memory and opt Graphics Cont
Hub Interface PCI Bus 4 EIDE Disks Super I/O Low Pin Count Interface
I/O Controller Hub (ICH)
Firmware Hub
PCI Controller IDE Controller DMA Controller USB Controller USB AC'97 Codecs (Box) Accel Hub Arch
Subsystems Subsystems in a PC communicate to each other via buses. Buses adhere to a particular architecture (set of rules) to allow compatibility with the numerous subsystems that adhere to the same architecture. Most PCs are associated with the term Wintel, which refers to Microsoft Windows and Intel chip technologies. Common Intel processors include Intel Pentium, Celeron, Pentium II, Pentium III, Pentium III Xeon, Pentium 4, Xeon and Itanium.
TXW02-R15
July 2001
32
Topic 1 - Introduction to Computer Hardware Architecture
PC Architecture: Controllers • All major subsystems have controllers • Circuitry controlling manner, method, and speed of access to device • Controllers are part of chip set Controller
Controls
Examples
L2 cache controller
L2 cache
256 KB
Memory controller
Memory
128 MB
Bus controller(s)
Data bus
PCI
Graphics controller
Monitor
S3 Trio64
Disk controller
Disk
40 GB IDE disk
Controllers All major subsystems have controllers that define how data will be obtained and stored. Sometimes a controller is a single chip with the data stored in separate physical circuitry. For example, a memory controller controls memory, but the data is stored in different physical chips called SIMMs or DIMMs. Sometimes a single physical chip contains multiple controllers. For example, it is common to have the memory controller and bus controller (PCI) in a single chip for desktop and notebook systems. Controllers are part of the chip set of the computer.
TXW02-R15
July 2001
33
Topic 1 - Introduction to Computer Hardware Architecture
PC Architecture: Buses Most transfers use three buses • Control bus • Address bus • Data bus
16-bit bus = 16 wires for on/off charges (data) 32-bit bus = 32 wires for on/off charges (data) 64-bit bus = 64 wires for on/off charges (data)
I/O Controller
Control
Data
Address
Processor
Data
Memory Disk
Graphics
LAN
• Some architectures multiplex signal on same bus (wires)
Buses If two subsystems are on a bus, such as in the diagram with processor and memory, a data transfer first involves sending the address on the address bus. Next, data is sent on the data bus. If multiple subsystems exist on a bus, a control bus is needed in addition to the address and data bus. The control bus is used to signal which subsystem will control the bus for the next transfer.
TXW02-R15
July 2001
34
Topic 1 - Introduction to Computer Hardware Architecture
Address Bus An address bus determines how much memory the processor or any subsystem can directly address. For example, a 32-bit address bus means 2 to the power of 32 or 4 billion unique numbers to address 4 GB of memory.
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
. . . . .
. . . . .
. . . . .
0 0 0 0 0
0 0 0 0 0
0 0 0 1 1
0 0 1 0 0
0 1 0 0 1
Memory Addressing Similar to Car Odometer
Before data is read or written by a processor, the address of that data is sent first. This address is sent on a separate set of physical wires called the address bus. The data is then sent on a different set of physical wires called the data bus. A processor is designed to use a certain maximum quantity of address lines. The amount of physical memory that a processor can address is determined by this quantity. The number of unique numbers that can be made by a base two number system (0s and 1s) with the quantity of address digits determines the maximum addressable memory of a processor. Software can limit this maximum addressability, for example, DOS sets the processor to use 20 address lines as DOS only addresses 1 MB of memory. Following are some processors and their addressability: Address lines
Addressable memory
Examples
24 32 36 44 64
16 MB 4 GB 64 GB 18 TB 18 EX
486SLC 486DX2 Pentium Pro, 460GX Itanium (Merced) IA-64 64-bit flat addressing
Sometimes operating systems limit addressability. For example, Windows NT 4.0 can see 4 GB of memory, but applications can only address 2 GB. Windows NT 4.0 Enterprise Edition lets applications address 3 GB.
TXW02-R15
July 2001
35
Topic 1 - Introduction to Computer Hardware Architecture
PC Architecture: Bus Speeds Processor + L1 Cache SCSI Bus 5 to 80 MHz
Front Side Bus 100 to 400 MHz
PCI Slots SCSI
AGP Graphics
ISA Bridge (Optional)
MCH or GMCH Host Bridge
I/O Controller Hub (ICH)
Memory System Bus if SDRAM: 133 MHz (speed of FSB) Memory Channel if RDRAM: 300, 356, 400 MHz
IDE Disks Super I/O
ISA 8 MHz
Backside Bus 600 MHz to 1.5GHz
AGP 66 to 133 MHz
PCI Bus 33 MHz (0 to 66 MHz)
L2 Cache
Firmware Hub
Hub Interface 133 MHz USB
Low Pin Count Interface 33 MHz
• Each bus is clocked at a different rate • Bus speed is different from data transfer rate (MBps) • Newer buses are double data rate (same MHz doubles throughput)
Bus Speeds Each bus in a PC has a speed (measured in megahertz) and a data transfer rate. In 1996 and 1997, the PC industry standardized around the 66 MHz Front Side Bus. Migration to a 100 MHz Front Side Bus occurred in 1998, then 133 MHz in 2000. The Pentium 4 introduced a 400 MHz Front Side Bus in late 2000. Data transfer rates (assuming that data is transferred on only one edge of the clock): • 8-bit at 133 MHz (double data rate) in 266 MBps (Hub Interface) • 32-bit at 33 MHz is 132 MBps (PCI bus) • 32-bit at 66 MHz is 264 MBps (PCI bus) • 64-bit at 33 MHz is 264 MBps (PCI bus) • 64-bit at 66 MHz is 528 MBps (PCI bus, frontside bus, and system bus) • 64-bit at 100 MHz is 800 MBps (frontside bus and system bus) • 64-bit at 133 MHz is 1066 MBps or 1.066 GBps (frontside bus and system bus; PCI-X) • 64-bit at 200 MHz is 1.6 GBps (backside bus to L2 cache; PC1600 DDR memory) • 16-bit at 400 MHz (double data rate) is 1.6 GBps (RDRAM channel)
TXW02-R15
July 2001
36
Topic 1 - Introduction to Computer Hardware Architecture
• 64-bit at 266 MHz is 2.1 GBps (PC2100 DDR memory) • 64-bit at 400 MHz is 3.2 GBps (backside bus to full speed L2 cache, Pentium 4 frontside bus) SCSI speed in megahertz is half of the rate measured in megabytes per second.
TXW02-R15
July 2001
37
Topic 1 - Introduction to Computer Hardware Architecture
PC Architecture: Cache • Cache is a buffer between subsystems • Disk transfer could involve 5 cache locations Processor + 1. L1 Cache
2. L2 Cache
MCH or GMCH Host Bridge
3. Memory
AGP Graphics PCI Slots
PCI Bus
4. SCSI
I/O Controller Hub (ICH)
EIDE Disks
5. Firmware Hub
Cache Cache: a storage place (buffer or bucket) that exists between two subsystems in order for data to be accessed more quickly to increase performance. Performance is increased because the cache subsystem usually has faster access technology and does not have to cross an additional bus. Cache is typically used for reads, but it is increasingly being used for writes as well. For example, getting information to the processor from the disk involves up to five cache locations: 1. L1 cache in the processor (memory cache) 2. L2 cache (memory cache) 3. Software disk cache (in main memory) 4. Hardware disk cache (common on SCSI controllers; EIDE usually uses only a FIFO buffer) 5. Disk buffer For reads, one subsystem will usually request more data than what is immediately needed, and that excess data is stored in the cache(s). During the next read, the cache(s) are searched for the requested data, and if it is found, a read to the subsystem beyond the cache is not necessary.
TXW02-R15
July 2001
38
Topic 1 - Introduction to Computer Hardware Architecture
Industry Standards Advanced Configuration and Power Interface (ACPI) • Operating system controls power management and configuration (plug-and-play) functions
PC 99 • Minimum requirements that a PC must meet to qualify for Designed for Windows logo • Systems compliant in 2H 1999 • No ISA devices or slots
PC 2001 • Will be implemented in late 2001
Industry Standards In the mid 1990s, power management for notebooks was handled by a standard termed Advanced Power Management (APM). This consists of power management hardware, a standardized (APM BIOS) interface between the platform BIOS and OS, and APM software. The APM system provides coordinated system power management, initially implemented in mobile PCs to aid in extending battery life. It was later implemented in desktop systems to reduce power consumption. In the late 1990s, control over power management moved from the BIOS to the operating system. This migration occurred as part of Advanced Power Management 1.2 and then the Advanced Configuration and Power Interface (ACPI) specification promoted by Microsoft and others. Advanced Configuration and Power Interface 1.0 was released in 1997 and is an open specification for easy and flexible power management to PCs; it enables the OS (instead of the BIOS) to direct power management. See www.teleport.com/~acpi/ for more information. ACPI defines a flexible and extensible interface that allows system designers to select appropriate cost/feature tradeoffs for power management. The interface enables and supports reliable power management through improved hardware and operating system coordination. The specification enables new power management technology to evolve independently in operating systems and hardware while ensuring that they continue to work together.
TXW02-R15
July 2001
39
Topic 1 - Introduction to Computer Hardware Architecture
The ACPI interface gives the OS direct control over the power management and plug-and-play functions of a computer. When it starts, the ACPI OS takes over these functions from legacy BIOS interfaces such as the APM BIOS and the PNPBIOS. The OS is then responsible for handling plug-and-play events and ensuring that power and thermal states are compliant with user settings and application requests. ACPI provides low-level controls so that the OS can perform these functions. So ACPI combines power management and plug-and-play. ACPI allows PCs to power on instantly and remain available to perform automated tasks after they are turned off. ACPI supports a low-power sleeping state that can be used instead of powering off the PC. It allows the PC to turn off peripherals such as CD-ROMs, network cards, disks, printers, and consumer devices attached to it such as VCRs, TVs, phones, and stereos. Connected devices can activate PCs also. Windows 98 and Windows 2000 implement ACPI 1.0. In Windows 98, it automatically installs itself and replaces APM; however, ThinkPad systems that have Windows 98 preloaded run in APM mode because of limitations of the current version of ACPI. ACPI has four power states for almost every component of a system. The four states, named D0 through D3, are a continuum from fully on to fully off with two levels of sleep modes in between. There are also five sleeping state definitions (S1 through S5), four processor power state definitions (C0 through C4), and multiple global system state definitions (G2, G3, and G5). Applications are not written to the ACPI calls (there is not ACPI-aware applications), so applications do not actively provide the operating system with detailed device usage requirements.
PC 99 and PC 2001 In August 1998, Microsoft and Intel released the PC 99 System Design Guide. The purpose of the guide is to ensure compatibility and interoperability among different brands of Windowsbased PCs. Hardware vendors must comply with the design guide to obtain a Designed for Windows logo from Microsoft to be used in advertising and marketing materials. The guideline requires the elimination of ISA slots by January 1, 2000 (PC 98 eliminated ISA systemboard slots); however, legacy systemboard implementation such as Super I/O is allowed. The guide is available at developer.intel.com/design/desguide. Microsoft and Intel introduced PC 2001 System Design Guide in late 2000 which systems must comply with to receive the Designed for Windows logo. Compliance begins in the second half of 2001 and is tied to the launch date of Windows XP.
TXW02-R15
July 2001
40
Topic 1 - Introduction to Computer Hardware Architecture
ACPI 2.0 In September 2000, a new version of ACPI was released as ACPI 2.0. ACPI 2.0 adds the support for 64-bit processor addressing to extend the benefits of ACPI to Intel’s IA-64 server architecture. The ACPI 2.0 specification also adds the concept of multiple processors and device performance states to enable longer battery life and lower operating temperatures on the platform, such as those currently supported by Intel’s mobile Pentium III processor featuring Intel SpeedStep technology. Enhancements targeted at the server market include support for hot-pluggable CPUs, memory, and PCI, PCI-X devices. The ACPI 2.0 was designed so that the ACPI 1.0 compliant portions of silicon would be compatible with ACPI 2.0. The complete ACPI version 2.0 specification is now available and can be downloaded from the ACPI web site at www.teleport.com/~acpi/. According to the participating companies, notebook PCs, desktops, workstations, servers and operating systems that incorporate ACPI 2.0 are expected to be developed by the end of year 2001.
TXW02-R15
July 2001
41
Topic 1 - Introduction to Computer Hardware Architecture
Benchmarks • Understand benchmark objective: either application throughput or subsystem performance • Examples include: - Business Winstone 2001 - WinBench 99 Disk WinMark 99 Graphics WinMark 99
- 3D WinBench 2000 - BAPCo (SYSmark) - ZD Content Creation Winstone 2001 - SPEC CPU2000
PC performance doubles every two years
Benchmarks Business Winstone 2001 is a system-level, application-based benchmark that measures a PC’s overall performance when running today’s top-selling Windows-based 32-bit applications on Windows 98 SE, Windows NT 4.0 (SP6 or later), Windows 2000, or Windows Me. Business Winstone doesn’t mimic what these packages do; it runs real applications through a series of scripted activities and uses the time a PC takes to complete those activities to produce its performance scores. It is a product of Ziff-Davis Media Benchmarks. WinBench 99 provides detailed information of all of the major subsystems of a PC--processor and memory graphics, disk, and full motion video. • Disk measurement is Disk WinMark 99; there are business and high-end categories. • Graphics measurement is Graphics WinMark 99; there are business and high-end categories. The WinBench 99 score gauges the performance of the CD-ROM subsystem, which includes the CD-ROM drive, its adapter, and its required software drivers. 3D WinBench 2000 evaluates 3D graphics performance for 3D-rendering features. BAPCo (Business Applications Performance Corporation) is a nonprofit corporation founded in May 1991 to create objective performance benchmarks that are representative of the typical business environment. IBM is a member. Common BAPCo benchmarks are SYSmark 2001 and WebMark2001. Contact www.bapco.com for more information.
TXW02-R15
July 2001
42
Topic 1 - Introduction to Computer Hardware Architecture
The Standard Performance Evaluation Corporation (SPEC) establishes, maintains, and endorses a standardized set of relevant benchmarks and metrics for performance evaluation. Contact www.specbench.org for more information. SPEC CPU2000 - Introduced in late 2000, SPEC (Standard Performance Evaluation Corp.) CPU2000 is a workstation application-based benchmark program that can be used across several versions of Microsoft Windows NT, Windows 2000, and Unix. It consists of two benchmark suites: SPECINT2000 which measures computation-intensive integer performance and SPECFP2000 which measures computation-intensive floating point performance. Both measure the real-world performance of a computer’s processor, memory architecture, and compiler. They replace CPUmark and FPUmark. ZD Content Creation Winstone 2001 - Content Creation Winstone is a system-level, application-based benchmark that measures a PC’s overall performance when running top, Windows-based, 32-bit, content creation applications on Windows 98, Windows NT 4.0 (SP6 or later), Windows 2000, or Windows Me. It is a product of Ziff-Davis Media Benchmarks. Ziff-Davis has several benchmarks available at www.pcmag.com/pclabs. Other Ziff-Davis benchmarks include • NetBench - Measures a server’s throughput for file read-and-write requests from clients. This is an effective benchmark for file/print servers. • ServerBench - Measures server response to client loads in a high-transaction environment. This is an effective benchmark for application servers. • WebBench - Measures server response speed to requests from web-browsing clients. This is an effective benchmark for intranet servers. • Business Winstone 2001 BatteryMark - Released in 2001, this benchmark provides a realworld approach to gauging the battery life of mobile systems. In 1988, the Transaction Processing Council (TPC) was formed to fulfill the need for transaction processing benchmarks that emulate the workloads found on database servers. The council includes representatives from a cross-section of 45 hardware and software companies that meet to establish benchmark content. A primary goal of the council is to provide objective and verifiable performance data to the industry. Visit www.tpc.org for more information. iBench 2.0 - The i-Bench 2.0 test suite is a comprehensive benchmark program designed to measure the performance and capabilities of Internet clients, using the latest Web technologies. The technologies found in i-Bench tests include HTML (Hypertext Markup Language) and XML (eXtensible Markup Language), Flash and Shockwave animation, Java and JavaScript processing, and various types of streaming and embedded audio and video. You can run iBench directly on the Web at www.i-bench.com on the Web.
TXW02-R15
July 2001
43
Topic 1 - Introduction to Computer Hardware Architecture
Server Performance Following is a listing of different server types and a ranking of the most critical subsystems that must be tuned to achieve optimum performance. Domain controllers • 1. Network I/O, 2. processor, 3. memory File and print servers • 1. Disk I/O, 2. network, 3. memory Application servers • Database servers – 1. Disk I/O, 2. memory, 3. processor • Communication servers – 1. Network I/O, 2. processor, 3. memory • E-mail servers – 1. Processor, 2. network, 3. disk I/O • Web servers – 1. Network, 2. memory, 3. disk I/O • Groupware servers – 1. Processor, 2. memory, 3. network For file and print servers (not application servers), the main goal of a network adapter is to move data quickly between the network and the disk subsystems of the server. The performance of the LAN adapter and disk subsystems can be seen in figure 1. As the number of users on the network or the server workload increases from zero, the performance of the network can be considered network intensive. Throughput or Transactions per Second
High Network Adapter Utilization and High Disk Cache Hit Rate
Poor Disk Cache Hit Rate High Disk Utilization
Low-End Network-Intensive
Midrange Processor & MemoryIntensive
High-End Disk-Intensive
Increasing Number of Users or Server Workload Figure 1. File Server Performance Characteristics
TXW02-R15
July 2001
44
Topic 1 - Introduction to Computer Hardware Architecture
IBM’s Netfinity 6000R achieved a solid throughput of 3,279 transactions per second in a hightransaction environment. Transactions per second
3500 3000 2500 2000 1500 1000 500 0 1
4
8
12
16
20
24
28
32
36
40
44
48
52
56
60
Client load Tests performed with Ziff Davis Media Inc.'s ServerBench 4.1
Throughput increases until it reaches a limiting peak and tails off as the slower disk subsystem becomes apparent. The more transactions that are processed, the more that writing to disk will be required. As the number of transactions increases, the performance moves into the diskintensive phase, which hinders the performance of the server. Figure 1 shows that by using a faster network card, the peak performance can be improved for low loads, but throughput will be restricted at higher loads by the disk subsystem. Figures 2 and 3 illustrate this issue. The use of a faster disk subsystem (number of disks, mechanical speed, and disk cache) can extend the peak throughput performance for higher server loads. The upgrading of both subsystems gives cumulative improvements. In addition, increasing the memory helps by increasing the file system cache. There is a common misconception that the processor is the most important part of the server and can be the single measure when comparing systems. These days, processors are so fast that their power is wasted; servers are often over-configured with processors and under-configured with disks, memory, and network subsystems. Only specific applications that are truly processor intensive take advantage of today's high-end processor.
Throughput (Transactions per Second)
Adding a Faster Network Adapter Benefit is Maximized During High Cache Hit Rate Benefit Drops Off as Disk I/O Rate Becomes the Bottleneck
Increasing Number of Users or Server Workload Figure 2. Effect of Adding Faster Network Adapter to the File Server
TXW02-R15
July 2001
45
Topic 1 - Introduction to Computer Hardware Architecture
Throughput (Transactions per Second)
Server With Faster Disk Subsystem
Server With Slower Disk Subsystem
Increasing Number of Users or Server Workload Figure 3. Effect of Adding a Faster Disk Subsystem to the File Server
TXW02-R15
July 2001
46
Topic 1 - Introduction to Computer Hardware Architecture
Summary: Computer Hardware Architecture • Notebooks, desktops, and servers are different types of computers with unique features • Information appliances have embedded operating systems and are easy and quick to install • Thin clients are useful for applications to execute on a server • Peer-to-peer computing uses information stored on other clients • Each computer subsystem is connected by • Wintel-based PCs adhere to industry standards such as ACPI • Industry benchmarks provide subsystem and total system performance comparisons
Computer Hardware Summary According to market research company Giga Information Group, information appliances will outsell personal computers by 2002. These information appliances will be used extensively both in small business and large enterprise environments.
TXW02-R15
July 2001
47
Topic 1 - Introduction to Computer Hardware Architecture
Review questions Objective 1 1. What type of system would market its size, weight, and docking station support? a. Notebook b. Desktop c. Server d. Information appliance 2. What type of system would most likely market its systems management support, chipset, and graphics controller? a. Notebook b. Desktop c. Server d. Information appliance Objective 2 3. Which statement best characterizes an Information Appliance? a. Requires embedded Linux operating system b. Ideal solution for enterprise mail function c. Requires a rack cabinet for installation d. Easy to install and add to infrastructure Objective 3 4. For the thin client known as Windows-based Terminal (WBT), where do applications and graphics execute? a. The applications execute on the server; the graphics execute on the terminal b. The applications execute on the terminal; the graphics execute on the server c. The applications execute on the server; the graphics execute on both the terminal and server d. The applications and graphics both execute on the terminal 5. Which statement is true regarding a thin client? a. A thin client is a desktop that is attached to a server via ethernet b. A thin client requires Remote Desktop Protocol (RDP) for connectivity c. A thin client requires Independent Computing Architecture (ICA) for connectivity d. A thin client can be a notebook or a desktop system Objective 4 6. Which of the following is not a peer-to-peer implementation? a. Central servers hold directories of files of registered users b. Central servers hold files of registered users c. User searches contents of shared directories of peer members d. Each PC is a mini-server for workgroup for file sharing and collaboration
TXW02-R15
July 2001
48
Topic 1 - Introduction to Computer Hardware Architecture
Objective 5 7. A device driver is an interface between what subsystems? a. The applications and BIOS b. Hardware and the operating system c. The operating system and the BIOS d. An API and the standard hardware
8. What circuitry controls the methods, manner, and speed in which a subsystem is accessed? a. Controller b. Keyboard c. VLSI logic d. Device driver 9. Most transfers of data between subsystems involve which buses? a. Data bus b. Data and address bus c. Data, address, and control bus d. Data, address, control, and tag bus Objective 6 10. What industry standard allows the operations system to control power management? a. Advanced Configuration Power Interface (ACPI) b. PC 99 c. PC 2001 d. OnNow 11. What is an important characteristic of a performance benchmark? a. Understanding if it measures a subsystem or application throughput b. The benchmark needs to incorporate Java c. Benchmark must include MPEG-2 encoding d. The PCI bus must be enabled
TXW02-R15
July 2001
49
Topic 1 - Introduction to Computer Hardware Architecture
Answers 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.
a. b. d. a. d. b. b a. c. a. a.
Notebook Desktop Easy to install and add to infrastructure The applications execute on the server; the graphics execute on the terminal A thin client can be a notebook or a desktop system Central servers hold files of registered users Hardware and the operating system Controller Data, address, and control bus Advanced Configuration Power Interface (ACPI) Understanding if it measures a subsystem or application throughput
If you have completed reviewing all topics of PC Architecture please return to the online course to take the CSAT survey and final test.
TXW02-R15
July 2001
50
