Windows Server 2003 Networking - 5

Choosing a file system

Windows servers provide three choices for the file system to format the server’s disk: FAT, FAT32, and NTFS. In most cases, you should elect to use NTFS. Well, actually, you should use NTFS in almost all cases. Come to think of it, you should always use NTFS.

The name FAT refers to the file allocation table that was used in the original version of MS-DOS back when disco was still popular. FAT was a simple but effective way to track disk space allocated to files on diskettes and on small hard drives. The original FAT system used 16-bit disk addresses to divide the total space on a disk into 65,526 units called clusters, each of which could be allocated to any file on the disk. The size of each cluster could vary from as little as 2K to as much as 256K, depending on the size of the drive. When disk drives started to get bigger than 512MB (can you remember when 512MB was a huge disk?), FAT was upgraded to FAT32, which used 32-bit addresses for clusters. That allowed a maximum of 524,208 clusters on the disk with the size of each cluster ranging from 4K to 32K, depending on the size of the drive. FAT32 was a huge improvement over FAT, but both suffer from several inherent problems:

✦ Even with 32-bit addresses, FAT32 is stretched by today’s 100GB+ disk drives.

✦ Neither FAT nor FAT32 have built-in security features.

✦ FAT and FAT32 are inherently unreliable. Most users, at one time or another, lose files due to the unreliability of FAT/FAT32.

✦ FAT and FAT32 allocate space inefficiently on large volumes because the smallest unit of space that they can allocate must be large — as much as 256K in some cases.

Here are just a few of the reasons that choosing NTFS for Windows servers is a no-brainer:

✦ NTFS has built-in security features that track security information for individual files and directories.

✦ NTFS tracks clusters with 64-bit disk addresses rather than 32-bit addresses (FAT32) or 16-bit addresses (FAT). As a result, an NTFS volume can theoretically have something in the neighborhood of 18 million billion disk clusters, which should keep you going for a while.

Installing and Configuring Windows 2000 and 2003 Server

Planning a Windows Server Installation

✦ The benefit of having so many clusters available is that the size of each cluster can be kept small. NTFS can efficiently use 4KB clusters for even the largest drives available today. ✦ NTFS drives are more reliable because NTFS keeps duplicate copies of important information, such as the location of each file on the hard drive. If a problem develops on an NTFS drive, Windows can usually correct the problem automatically without losing any data. In contrast, FAT drives are prone to losing data.

✦ The system that FAT uses to keep track of which disk clusters belong to a given file is prone to errors. In contrast, NTFS has more redundancy built in to its record keeping, so it’s less likely to scramble up your files.

✦ NTFS has better support for large drives and large files. Table 1-2 compares some of the maximums of the FAT, FAT32, and NTFS file systems.

Table 1-2 File System Limits

Planning your partitions

Partitioning enables you to divide a physical disk into one or more separate units called partitions. Each disk can have up to four partitions. All four of the partitions can be primary partitions, each of which can be formatted with a file system, such as NTFS or FAT32. Or, you can create up to three primary partitions and one extended partition, which can then be subdivided into one or more logical drives. Then, each logical drive can be formatted with a file system. Although you can set up partitions for a Windows server in many ways, the

following two approaches are the most common:

✦ Allocate the entire disk as a single partition that will be formatted with  NTFS. The operating system will be installed into this partition, and disk space that isn’t needed by the operating system or other network applications can be shared.

✦ Divide the disk into two partitions. Install the operating system and any other related software (such as Exchange Server or a backup utility) on     the first partition. If the first partition will contain just the operating system, 10GB is a reasonable size, although you can get by with as little as 4GB if space is at a premium. Then, use the second partition for application data or network file shares.

Deciding your TCP/IP configuration

Before you install the operating system, you should have a plan for how you will implement TCP/IP on the network. Here are some of the things you need to decide or find out:

What is the IP subnet address and mask for your network?

✦ What is the domain name for the network?

✦ What is the host name for the server?

✦ Will the server obtain its address from DHCP?

✦ Will the server have a static IP address? If so, what?

✦ Will the server be a DHCP server?

✦ What is the default gateway for the server? (That is, what is the IP address of the network’s Internet router?)

✦ Will the server be a DNS server?

If the server will host TCP/IP services (such as DHCP or DNS), you’ll probably want to assign the server a static IP address.

Choosing workgroups or domains

A domain is a method of placing user accounts and various network resources under the control of a single  directory database. Domains ensure that security policies are consistently applied throughout a network and

greatly simplify the task of managing user accounts on large networks.

A workgroup is a simple association of computers on a network that makes it easy to locate shared files and printers. Workgroups don’t have a sophisticated directory database, so they can’t enforce strict security. Microsoft says that workgroups should be used only for very small networks with just a few users. In fact, any network that is large enough to have a dedicated server computer running Windows 2000 Server or Windows

Server 2003 is too large to use workgroups. As a result, if you’re installing a Windows server, I recommend that you should always opt for domains. After you decide to use domains, you have to make two basic decisions:

✦ What will the domain name be? If you have a registered Internet domain name, such as mydomain.com, you may want to use it for your network’s domain name. Otherwise, you can make up any name you want.

✦ What computer or computers will be the domain controllers for the domain?

If this is the first server in a domain, you must designate it as a domain controller. If you already have a server acting as a domain controller, you can either add this computer as an additional domain controller, or designate it as a member server. You can always change the role of a server from a domain controller to a member server and vice versa if the needs of your network change. For small networks, I suggest that you use just one domain controller and designate the other servers as member servers. For larger networks, you may want to create several domain controllers to spread the work around.

Before You Install . . .

After you’ve made the key planning decisions for your Windows server installation, you should take a few  recautionary steps before you actually start the Setup program. The following sections describe the steps that you

should take before you perform an upgrade installation. Note that all these steps except the last one apply only to upgrades. If you’re installing a Windows server on a new system, you can skip the first steps.

Backing up

Do a complete backup of the server before you begin. Although Windows Setup is reliable, sometimes  something serious goes wrong that results in lost data. Note that you don’t have to back up the drive to external media, such as tape. If you can find a network disk share with enough free space, back up to it.

Checking the event logs

Look at the event logs of the existing server computer to check for recurring errors. You may discover that you have a problem with a SCSI device or your current TCP/IP configuration. Better to find out now rather than in

the middle of Setup.

Uncompressing data

If you’ve used DriveSpace or any other disk compression software to compress a drive, you’ll have to uncompress the drive before you run Setup.

Neither Windows 2000 Server nor Windows Server 2003 support DriveSpace or other disk compression programs.