Note

The disk space required after setup depends, of course, on the amount of RAM on the system and thus, of the size of the paging file.

It goes without saying that you won't install servers that only meet minimum requirements. In fact, if you're planning on putting together an enterprise network, they won't be at Microsoft's recommended levels either. If you're wise, you'll either simply double Microsoft's recommendations and use that as a starting point or perform a formal Server Sizing Exercise. This exercise will help you determine the hardware and software configuration for each of your servers. It will tell you what size server you need, where it is needed, and what it should deliver in terms of services. When configuring servers, don't forget to take the following items into consideration:

• Identify server bases Identify where your client groupings are. You will need to position your servers where you have a concentration of clients or users.

• Number of users per server Identify a maximum number of users per server. To provide a given level of service, you need to ensure that there are never more than the specified number of users, depending on this server's services. On average, organizations set up one server per 250 users, but this depends on the server's function because with WS03, servers can support thousands of users.

• Maximum acceptable server load Determine the speed of response you want from a server when providing a given service. This load must take into consideration the maximum number of users as well.

• Server variance The location of the server is also important to consider because it often serves to determine the nature of the server. Most servers located at headquarters or in large regional offices will tend to be single-purpose servers—they will either perform one role or another. Servers in smaller regional offices, on the other hand, are often multipurpose servers.

If a regional office has fewer users than the minimum number of users per server that you determined earlier, more than one server would be too costly and will rarely be budgeted. So if you have only one server and you have a series of different services that must be delivered, you need to configure a multipurpose server. Multipurpose server configurations will differ from single-purpose servers because they are isolated. As such, they often need to be independently recoverable.

• Minimum server capacity Determine the minimum hardware capacity you want for your servers. Remember that you don't want to change them for some time. The purpose of your network is to deliver services to your user base. Like most people, you'll want to provide a quality service. Take this into consideration when you determine the minimum server capacity. Capacity planning should identify items such as number and size of the processors, amount of RAM, and disk size. Each item is influenced by the decisions you've made before: How many users will the server cover? Where will the server be located? Will it be single or multipurpose?

• Multiprocessing In most cases, you will use multiprocessing servers, servers that have more than a single processor. You'll have to take care here, since there is a clear demarcation between multiprocessor systems. The Standard Edition supports only four processors. All systems with five to eight processors require the Enterprise Edition. This will have an impact on your server budget.

• RAM sizing The rule is simple: the more RAM you have, the better your server will perform. Thus, RAM is not an item you should skimp on. It all depends on the function of any given server, but it is a good rule of thumb to double Microsoft's minimal recommended requirements and start all servers at 512 MB of RAM, then go up from there. Use RAMBUS technology since it is a lot faster than EDO, DDR, and SDRAM and is becoming more comparable in pricing.

Some server functions are RAM-intensive, such as Terminal Services or Application Servers. These will require more than the minimum you set. In addition, RAM size affects the paging file. The best practice here is to start the paging file at double the size of your RAM and set its maximum size to four times the size of RAM. This rule changes when you're dealing with massive amounts of RAM such as 4 GB configurations, but at first, it means that you'll need to reserve a minimum of 2 GB of disk space for the paging file.

• Disk sizing The size and number of disks you put into each server will depend on a number of factors. How many partitions do you want to make? How much space do you want to reserve for the operating system, programs, and special elements such as the paging file? How much space for data storage? Most servers will end up with three partitions: one for the server utilities, one for the operating system and programs, and one for data. Windows Server 2003 uses only the last two partitions. The operating system partition should also store the paging file. Keep in mind that WS03 offers a better performance when it reads and writes to multiple disks, so you might want to reproduce the paging file on other disk drives. If that is the case, each drive will need to reserve the same amount of space for this file. System drives should be a minimum of 4 GB and should be more if you plan on having a lot of RAM in your server, because it will affect the size of the paging file.

Data partitions should always be separate from system partitions and are most often significantly larger. Keep in mind that if you are preparing a file server to store user data, you'll have to offer a valid storage size on a per user basis. Many organizations don't have a consistent storage policy. They offer 50 MB of storage per user, something almost no one can live with today, and they insist that any data stored on the user's local PC is not protected by the organization. If you plan on storing user data, you'll have to consider allocating at least 200 MB per user and expect that it may well grow to 1 GB per person. It all depends on the type of activity they perform. But, worry not, disk space is a lot cheaper today and is always becoming more so.

• Hardware protection All this data needs some level of protection. Local disk drives should be protected by a random array of inexpensive disks (RAID). Many people opt for a disk mirroring system (RAID 1) for the system drives and stripe sets with parity (RAID 5) for data partitions. There are differing opinions, but with today's fast-paced advances in disk technology, it is quite acceptable to opt for a single RAID 5 system and partition it into two for system and data drives. Don't forget the RAID overhead: 50 percent more disk space is required for RAID 1 and a minimum of 20 percent is required for RAID 5. This is 33 percent if you have the minimum number of drives to support RAID 5 (three drives).

You can also use a random array of inexpensive network (RAIN) cards. They are similar to a RAID disk system in that they are composed of two network cards using the same resources. When one fails, the other automatically takes over using the same MAC address.

• Storage strategy The hardware protection system you choose will also depend on your storage strategy. If you're building a multipurpose regional server, you'll probably want to focus on local storage. Thus, you'll design a suitable local RAID solution. But if you decide to centralize storage for single-purpose servers, you'll want to implement a storage area network (SAN). In this case, you'll need to consider storage requirements for all servers at once and change your strategy for operating system storage. In fact, WS03 servers can even boot from a SAN, letting you create diskless server configurations.

• Physical location The physical location, the actual physical space the server will occupy, will help you determine whether you will choose a rack-mounted or tower server configuration. In most cases, multipurpose servers are tower servers and single-purpose servers are rack-mounted because they are concentrated in a single physical space. This physical location should be lockable and offer temperature controls, and all physical access to servers should be audited.

• Backup method Once again, the physical location of the server will help determine the backup method selected. Regional servers often use tape drives for backup, but this depends on the speed and available bandwidth of your wide area network connection. Central servers use remote backup solutions such as tape or writable DVD robots. This solution can service regional servers as well if the appropriate network bandwidth is available.

Time will also be a factor in this decision. If you choose a technology that cannot back up the system in the amount of time that is available, you'll be creating a problem, not solving it. Windows Server 2003 helps here since it has the ability to do backup snapshots—time-based images of the hard disk drives that are then used to create the backup, allowing the server to continue with other operations. More on this topic will be covered in Chapter 9.

• Operating system Are there any special requirements for the operating system this server will host? For Windows Server 2003, it's easy. Everything—hardware and software—has to be certified. Microsoft has made great advances in stability with its operating systems, but these advances depend on products that follow strict guidelines. In an enterprise network, only certified products are allowed. If you have existing hardware that is not certified, you'll have to weigh the risk of using it on a critical component such as a server against the cost of buying replacement parts. If you're buying new hardware or software, make sure it is certified for Windows Server 2003.

• Growth potential Finally, you don't want to be replacing this system six months after you put it in place, so make sure that it has a lot of capability for growth. All systems should have the ability to add more processors, more memory, and more disk space. As such, you'll need to consider the server life expectancy—when the server was introduced by the manufacturer, when it will be retired, its projected growth potential by the manufacturer, and so on. If you plan carefully, you'll be able to implement servers that will have a rich lifecycle that will meet your expectations. In some conditions, this lifecycle can last up to five years.

This exercise helps you identify the generic size of each server. Special functions such as domain controllers or Microsoft Exchange Servers will require different sizing parameters. Microsoft offers sizing tools for most of its .NET Enterprise Servers family. All are available on the Microsoft Servers Web site at http://www.microsoft.com/servers/. In addition, Dell, Hewlett-Packard, and IBM all offer sizing tools for their servers on the appropriate Web sites.

Upgrading Existing Systems

If you already have a network in place and you want to upgrade to WS03, you'll first want to check if your systems are going to be compatible to WS03. Like Windows 2000, the Windows Server setup includes a special feature that will verify if the existing system can be upgraded to WS03. Keep in mind that upgrades will only work with the Standard, Enterprise and Datacenter Editions of Windows Server 2003.

To perform a verification, insert the appropriate Windows Server 2003 CD and select Check system compatibility from the Startup menu. Next, click Check my system automatically. Windows Setup will offer to download the latest drivers and updates from the Windows Update site. If your system is connected to the Internet, it is a good idea to select this option. To do so, select Yes, Download the updated Setup files. If you do not have an Internet connection, click No, Skip this step and continue installing Windows.

Was this article helpful?

0 0

Post a comment