An operating system must be able to anticipate future growth of hardware and components. One area that has been continually improving is storage. Hard disk and RAM capacities are increasing at an alarming pace. Luckily, Windows NT supports a large amount of RAM (4GB) and an even larger capacity of hard disk space (16 exabytes.) Windows NT also can support removable media such as SyQuest, Bernoulli, and other magneto-optical devices. These drives are often used as offline devices for data archiving to tapes or compact discs, but also can be used for online storage in addition to the hard disks. The removable media are not quite as fast as hard disks, but they can be removed and placed in other devices, or stored offline for later use.

With your hard disk drives you can enable the use of striping, which spreads the data among several (a minimum of two) hard disks in the system, which can be read by the hard disk controllers at once. This greatly increases disk I/O performance. It is not a fault tolerant method, and a failure of one of the stripe set members results in a loss of the entire stripe set. You also can configure the hard disks in your system to use a volume set. A volume set takes areas of free space on any drive of a partition and creates one logical drive. This is useful when you have areas of the disk that are too small to be worthwhile. Once again, this is not a fault tolerant feature. Both of these methods will be discussed in later chapters.

The Windows NT architecture can be daunting, but an understanding of how and why NT works is critical. Refer to Figure 1-1 as you study this topic. It is interesting to see how many components are involved. Windows NT Workstation is designed as components that are responsible for their own functions. These modular components comprise the NT architecture. The major components of the architecture are the Hardware Abstraction Layer (HAL), the Kernel, the Executive (System Services), and the Environment subsystems. How a process accesses the hardware is determined by what mode the process is running in: User Mode or Kernel Mode.

Most NT code is stored and most applications run in User Mode. This is also the mode in which the environment subsystems that we discussed earlier are executed. User Mode is designed to prevent applications from bringing down the operating system. If a process is running in User Mode, it cannot access the hardware directly. It must call upon the operating system to access the hardware. This provides stability for the operating system by providing a protective barrier against a misbehaving application. User Mode is also called Nonprivileged Processor Mode.

Kernel Mode, also called Privileged Mode, has direct access to hardware and software resources in the system. Calls for hardware functions from applications running in User Mode must be translated into Kernel Mode instructions before the hardware can be accessed. The data must then be retranslated into User Mode when it is returned to the application. An important note about the release of NT 4 is that some components of NT that used to run as User Mode components now run as Kernel Mode components. These are the Window Manager, GDI, and graphics device drivers.

The Hardware Abstraction Layer separates the Kernel from the hardware to provide an intermediary layer, so that the Windows NT Kernel does not have to perform communication with the hardware. You should understand that the User Mode applications never call the hardware directly. These calls are handled only by the Windows NT Executive. Because of this, HAL is the most common reason why DOS and Windows applications are incompatible with NT. These applications expect to communicate directly with the hardware, and will not run if they are not given this privilege. Not only do you need another HAL for each hardware platform, you need two for each processor architecture. One is for supporting a single processor, and the other is for supporting multiple processors.

The Windows NT Kernel is the most important portion of the NT architecture, and is designed to be small and efficient to handle operating system events. This is the core of the operating system. The Kernel is small because the other components of NT provide their own functions and leave the Kernel to perform its duties. Additional components can be added without affecting or disrupting the Kernel itself. This is important when porting Windows NT to other processor architectures. One of the most important duties of the Kernel is to manage the use of threads, which can be defined as units of execution. Not only does the Kernel manage the threads, it is responsible for creating and scheduling them. These threads are assigned a priority from 0 - 31. (A higher number means a higher priority.) The Kernel must send threads to be processed based on this priority, or preempt a thread that has used up the time slice given to it.

The Executive (System Services) provides the operating system services that the Kernel is too busy to perform. The Kernel is actually part of the NT Executive, and together they are responsible for the entire system. The NT Executive is responsible for numerous major components.

The Environment subsystems provide support for the various application types that can be run, such as POSIX, Win32, and OS/2. Such applications require a particular environment, which is mimicked by the Environment subsystem. Environment subsystems are considered subsystems, because they can perform independently of other subsystems. They are described as protected, because other subsystems do not have the right to access another application’s subsystem. Each of the major subsystems is described below:

The primary subsystem for Windows NT, Win32 is used for all user input and output, and is called upon by the other subsystems for services. An interesting note is that the OS/2 subsystem originally was going to be the primary subsystem for Windows NT, until the surprising success of Windows 3.x.

MS-DOS and Win16 were included within Windows NT to provide backward compatibility with older DOS and Windows applications. Most applications being used in their native operating environments now can be used in NT. The use of the NTVDM enables these applications to run properly. This is described in detail earlier in the chapter.

The POSIX subsystem was included to support open standards, mostly for application support for UNIX platforms. The POSIX subsystem is discussed in greater detail in later chapters on file systems.

The OS/2 subsystem was originally planned to be much more significant in NT, but dwindling support has decreased the need for OS/2 compatibility. The OS/2 subsystem is protected and can be preemptively multitasked, and also can translate the Win32 calls. However, the OS/2 subsystem does not support graphical applications. It can support only Character Mode applications, which are scarce. Therefore, the OS/2 support is very limited.

Windows NT has made significant improvements in its memory architecture. The most important features that comprise the NT memory architecture are as follows:

Virtual Memory allows your NT Workstation to use the hard disk to simulate RAM as needed. NT uses a system called paging to move an unused portion of memory to the hard disk, and to retrieve the data when it is needed. This also can be called demand paging. NT accomplishes this through use of the Virtual Memory Manager, and the paging file. It is VMM’s job to determine which data has been paged to the disk, and to keep track of the physical memory addresses used by this data. The paging file is the actual file on the hard disk where this information is paged to and from. You will learn more about the paging process and configuring the paging file for use with NT in later chapters.

The 32-bit flat address space is used in Windows NT instead of the segmented memory architecture used in MS-DOS and Windows 3.x. Segmented memory architecture used a 16-bit addressing model to access memory locations using a segment/offset method. (This method is akin to a map grid. You find the location of F7, for example, by moving over to column F, and down to row 7. The segment and the offset are comparable to columns and rows.) It took valuable system resources to map these memory references. The 32-bit flat address space means that applications can access up to 2GB of RAM, rather than the 64K segments that were previously used. This is what other non-Intel processors use as their address scheme, therefore NT is more compatible with these processor architectures.

Windows NT represents the pinnacle of desktop operating systems to date. Its power, compatibility, security, and ease of use make NT Workstation, quite possibly, the best workstation operating system. It’s more stable than Windows 3.x or Windows 95. It’s easier to use than the SPARCstation. And it’s more compatible than OS/2. Since Microsoft dominates the market as far as desktop operating systems go, you will most likely be choosing among various Microsoft operating systems in your organization. You must understand the features of each operating system in order to make a wise decision. There is nothing wrong with using a combination of desktop operating systems in your organization. This can be an effective way to cater to the capability and purpose of the computer. There are ramifications to implementing a mixed network, mostly involving compatibility. If your organization uses certain applications throughout the company, each platform has to be able to support those applications. (For example, if you use Microsoft Office for Windows 95 or NT in your company, how could the 16-bit operating systems, such as Windows 3.x, share documents and worksheets created in these versions? Workstations using an older version of the Office suite couldn’t read the formats created using the Windows 95 or NT versions. The newer version of Office could read the formats of older versions, but it doesn’t work the other way.) These considerations must be weighed when making operating system decisions. You should plan an extensive period for testing in a lab environment. Install the new operating system, and all the applications you plan on using. Determine if there are any incompatibilities before the upgrade. If there are, will you upgrade to a newer version? Discontinue the use of the application? Find another, equal application from another vendor that does support your new operating system? Or just not upgrade to the new operating system?

The choice between NT Workstation and NT Server depends on the use of the machine. If you use the machine as a source for applications, files, or print services, it would be wise to choose NT Server over Workstation, because NT Server is optimized for these types of services. That is not to say that you cannot share files or print devices with NT Workstation. You can share these services just as you would with NT Server, but you might decide between the two based on how many of these services a given machine shares at one time. If other users often log on locally to the system, you might elect to use NT Workstation. (Logging on directly into NT Servers should be reserved for administrators qualified to support NT Server.) If you need services available for sharing at all times, however, choose NT Server. Any processing required for a user sitting directly at the system detracts from the capability to share resources such as applications and files. However, you can administer the domain from a workstation without the need to log on locally at the server. Table 1-1 shows a detailed breakdown of the various differences between the two operating systems.

Other than the features just listed, the two operating systems are very much alike. One of the most important differences mentioned in Table 1-1 is the number of inbound client sessions available. The limit of 10 client sessions to your NT Workstation is a drawback if you are using your workstation to share files or applications. If you are constantly reaching this limit, transfer these services over to an NT Server so all users can access the resources in a timely manner. After all, that is the purpose of resource sharing: to give users access to files, printer, and applications. The other important difference is that NT Server is optimized for network throughput and NT Workstation is optimized for better local access.

The hardware requirements of the two operating systems are quite similar. This is, of course, the bare minimum. It would be wise to increase the amount of RAM on the NT Server to 32MB or more, to increase response time. Even more than 32MB of RAM is recommended if your NT Server is also running any of the BackOffice products.

The limitations of NT Workstation when it comes to fault tolerance are also important. Although you can stripe the data on NT Workstation to increase the disk I/O, there is not an option for disk striping with parity. This is a feature supported in NT Server that enables you to recreate the data in the stripe set if one of the members of the stripe set fails. NT Server uses the parity information to recreate the data. Another fault tolerant feature that NT Workstation does not support is the capability to mirror drives. NT Server can mirror the information from one drive to another to eliminate the possibility of data loss if one of the drives fails. Although these features would be useful for some users in NT Workstation, their absence does not pose a big concern. If you participate on an NT network, you should save all of your important information to the server, where the data will be backed up during the regularly scheduled backup routine for the servers. If your workstation hardware fails, you need only replace the failed devices, in this case a hard disk, reinstall NT and your applications, and you are back up and running.

Your choice between Windows NT Workstation and Windows 95 for a client workstation depends on your working environment. For the average home user, Windows 95 is the logical choice. This might not be the case in special situations, so I’ll discuss the differences between the two operating systems in this section.

There are many applications and programs available for the Windows 95 operating system. Most of these will run on NT Workstation. When run in Windows NT, they might receive a performance increase. We have learned earlier that NT runs DOS applications in their own subsystem, and preemptively multitasks the VDM. This results in a performance increase. Windows applications benefit in terms of greater stability, because they can be run in their own separate memory space. This prevents an errant application from bringing down other applications sharing the same space, as often happened in their native operating environments. If you are using your home system for games, you will find a few that refuse to run in NT. NT is not the preferred game machine, due to the fact that some of these games expect to have access to the hardware. This can be remedied by dual-booting between Windows 95 and NT. On my own computer, I have to dual-boot between the two operating systems because of a scanner that is not supported under Windows NT.

Windows 95 would also be used on systems that do not meet the hardware requirements of NT, or devices in the system that do not adhere to the strict Hardware Compatibility List provided for NT. Windows 95 runs on a 386/20 with 4MB of RAM, although very poorly. Windows NT requires at least a 486DX/33 with 12MB of RAM. If you require plug and play functionality, you have to choose Windows 95 for now, because NT is not plug and play-compatible as of yet. Our company uses Windows 95 on the laptops, because so many users complained about sluggish performance by NT, and the lack of power management.

Home users also don’t have much use for the added security features that NT provides. Its capability to restrict file and folder access for users or groups probably would never be used. Neither would its capability to audit successful and failed attempts to access system resources.

The new NTFS file system offers home users and business users considerable advantages over the FAT file system of Windows 95.

As we mentioned earlier, the added security with the NTFS file system might not mean much to home users, but in a business setting this is a very valuable feature.

Windows 95, unlike NT, doesn’t support the creation of stripe and volume sets—features that any user would find beneficial.

Table 1-2 shows the differences between Windows 95 and Windows NT.

Another distinction between the two operating systems is the cost. Windows NT Workstation costs more than Windows 95. If you feel you have the computer to support NT and are aware of the factors involved with your decision, the higher cost for NT might be a worthwhile investment in terms of performance and features.

Windows NT supports the use of a workgroup or domain for the managing and sharing of resources for users. In any network of computers, resources are shared, and clients access the shared resources. Each of the models, workgroup and domain, supports the sharing of resources, but they do so in different ways. The choice of workgroup or domain is based on the network environment and several other factors. Important factors to consider in deciding between the two are the number of users who will share the resources, and their technical knowledge.

The workgroup model, illustrated in Figure 1-2, is most often used in a small group of users, due to the way resources are managed and shared. Each computer in the workgroup must maintain the account information for each user in the workgroup. This is because there is no central means of providing authentication for users. If you were to change your account information, you would have to change your account information on every computer that you access in the workgroup before you can access resources. This gets more complicated as more users are added to the workgroup. If your workgroup consists of fewer than ten users, the workgroup model provides an effective way of sharing information, provided all the users who participate in the workgroup have the technical ability to share system resources and manage account information. This is the most complicated facet of the workgroup environment, but the workgroup model is much easier than the domain group model in terms of user account management and access to resources.

Under the domain model, a Windows NT Server acts as a domain controller. The domain controller authenticates users into the domain before they can access resources that are a part of the domain. See Figure 1-3 for an illustration. These resources can be located on the NT Server, or any computer on the network. The domain model is strongly recommended for networks of ten computers or more. The domain model is also recommended, regardless of the number of participating computers, if you require central control of user management or resource access. For this reason the domain model is by far the model of choice for businesses. It enables an administrator to centrally manage the accounts in the domain, including account policies that regulate password restrictions such as length of password, password age, and account lockout. The administrator also can add users to groups, give the user special permissions, and disable or delete accounts from the Windows NT tools including User Manager available in NT Workstation, or User Manager for Domains in NT Server.

As we indicated earlier, in the workgroup model user account information is kept on each computer that participates in the workgroup. In the domain model, the user account information database can be shared by users or groups, and also is shared by the servers in the domain. These servers require access to this database if they are to perform logon validation for users. They are, most likely, the Primary Domain Controller or a Backup Domain Controller. You will learn more about these types of servers in later sections of the book.

The domain model has a few variations when you are using more than one domain, as in the case of large enterprise networks with separate divisions or departments. Domains can be organized as physical or logical divisions within a company. For example, a domain model based on physical locations would include domain names such as Buffalo and Seattle. An example of a domain model based on logical groupings of computers would be domain names such as Accounting and Sales. Either way, the domain should be named to illustrate its purpose. You will learn more about domains and the various domain models as you continue towards your MCSE with the Windows NT Server exam, and especially the Windows NT Server in the Enterprise exam. For now, we are just clarifying the differences between the workgroup and domain models.

Figure 1-3 can be a little misleading. We have stated that a domain controller is needed to authenticate users upon logging on to the domain. Why are there two NT Servers in the illustration? One of the servers is the Primary Domain Controller (PDC), which is at the top of the NT domain hierarchy. The PDC is responsible for the changes to the directory database. Any adjustments by the administrator are recorded in this database on the PDC. There would be one PDC for each domain. The other computer in the illustration is a Backup Domain Controller (BDC). The BDC maintains a copy of the directory database that has been replicated from the PDC. The BDC can authenticate user logons with this database. Therefore, any of the domain controllers can be used for authenticating the client workstations.

Now that you have seen the two models, here are possible scenario questions relating to the topic: