Once you have decided how to organize your network and picked an NOS, you need to select the hardware for your network server. Your organization will depend on this system for critical business functions and you want to be sure it offers adequate performance and reliability-without breaking your budget. It is possible to spend tens of thousands of dollars on a network server, but a more modest investment will likely suffice in a small network. Here are the important points to consider:

Select a server-class computer. While desktop computers are less expensive, they lack some of the features needed for full-time server operations.

Processor type. Small networks almost always use Pentium-based servers. If your organization favors Macintosh computers, then a PowerPC-based server will work well as an AppleShare server, but it does not run Windows NT or NetWare. While earlier versions of NT ran on the PowerPC, Microsoft no longer develops new versions of NT for this platform. There are also RISC processor systems to choose from such as Sun's UltraSPARC. While versions of NT that will run on this platform exist, most small business environments will stay with the Intel architecture because it is simpler and more familiar.

Processor speed. Select the fastest processor that you can afford. Faster processors come on the market with increasing frequency. The primary tradeoff is between price and performance. Buying a server based on the very latest and fastest processor will cost your company significantly more than the previous generation. If the budget for your network is tight, then purchasing the second-fastest class of processor can achieve significant savings. Otherwise, purchasing the fastest processor available will extend your investment by allowing you to make use of that server for a longer time before it becomes underpowered.

Number of processors. One of the distinguishing features of server-class computers is the ability to support multiple processors. Advanced operating systems such as NetWare, Windows NT Server and Solaris can perform computing tasks faster using multiple processors through symmetrical multiprocessing. The number of processors a server needs depends on the number of users it supports, and the intensity of the work. In large networks supporting thousands of users, it is not unusual for file servers to be equipped with four or more processors. For medium-scale applications a single high-performance processor will handle 25 users with adequate performance. As a general guideline, most small business networks can easily rely on a single-processor server. A common configuration available from many computer vendors are dual-processor systems that arrive with only a single processor installed. These allow you to operate with a single processor now and add a second processor later.

Memory. An OS thrives on a memory-rich system. Servers rely on memory not only for running applications, but also for data caching. For either NetWare or NT, install at least 64 MB of memory for a basic file server, and even more if you will be running other applications as well. Adding memory is one of the most cost-effective means to boost server performance. As your network operates over time, closely monitor how it uses memory and add more if the system becomes constrained.

Disk architecture. While IDE drives dominate most desktop computers, SCSI drives dominate server-class computers by virtue of their superior performance. The SCSI architecture makes it relatively easy to add drives or other storage devices such as tape drives and CD-ROMs.

Your file server's disk system is an extremely critical device in that it holds one of your organization's most valuable assets-its data. All reasonable measures must be taken to ensure that data cannot be lost in the event of a hardware failure, power outage, software malfunction or human error. One option for the disk storage in a file server is to use RAID (redundant array of inexpensive devices). RAID-based storage systems use several SCSI disks configured in such a way that even if one drive fails, the storage systems continue to function with no loss of data. To achieve this ability to survive hardware failures, a RAID system will have more physical disks than the amount of available storage. A significant part of the space on the disks stores parity information, which can be used to reconstruct the data on the other disks in the event of a failure. When a disk fails, data are derived from parity, and when the failed disk is replaced, its data is reconstructed. Most RAID systems will also include redundant power supplies, and software to configure and monitor the system. Most rely on a specialized disk controller. RAID systems cost significantly more than raw disk storage. The factors that would motivate an organization to use RAID would be the cost of data loss and the impact of downtime. If the nature of your environment is such that you can live with a worst-case scenario of losing one day's or part of a day's worth of data, then RAID-based storage might not be necessary. Again, small networks are the ones least likely to require an industrial-strength storage solution based on RAID.

Data backup. It is standard practice in data networks to regularly make copies of all data. Most networks rely on tape as the backup media, but devices such as optical and Jaz drives can be effective-especially for small networks. Back up your data at reasonable intervals, but do it frequently. A standard approach would be to back up all data once a week, and all new and changed files daily. This strategy ensures that no more than one day's worth of data would be lost in the event of a major failure. If the nature of your network is such that even this degree of vulnerability is too high, then increase the frequency of incremental backups to multiple times per day, or invest in a RAID storage solution.

There are numerous options for the type of media you can use for backups and how to integrate it into your network. The most common media for archiving data for backups is 8mm tape. Most 8mm tape drives are SCSI devices, and can be integrated into the file server itself or purchased as a separate peripheral.

It is just as important to have the right software for data backups as it is the right hardware. While it is possible to perform backup tasks manually, automating the process is highly desirable. The consequences of forgetting to perform a backup can be quite high. There are several automated backup systems on the market, and often a tape drive will be included with this type of software. Here are some of the features to look for in a backup system:

• The ability to automatically schedule full and incremental backups.
• Reports of all backup activities, and notification of any failed or incomplete tasks.
• Verification of archived data.
• Compression of archived data.
• Support for all file types. vAbility to archive files that are in active use.
• Tracking of all archived files.
• Ability to selectively restore files.
• Ability to back up files on multiple servers and multiple kinds of servers (NetWare, NT, Unix).
• Ability to back up files on client workstations.