1. Web servers - these provide the basic pages and not so basic pages associated with the internet
2. E-Mail servers - provide the basis of e-mail
3. File servers - provide files for a large variety of purposes over the Internet
4. Storage/Backup servers - provide storage and or backup for particular information

Servers are almost always totally different than desktop computers.  They are designed with different purposes in mind.  It is almost always very important to purchase computers that are fundamentally different from desktops when purchasing servers.  Some of the major features include

Some of the major features of servers include

• More internal space - servers usually have more physical space inside for devices such as hard drives and slots for cards and CPUs.
• More CPU performance - servers can have multiple CPUs or very fast single processors
• More upgrade options - servers usually have more upgrade options than desktops.  They are designed for growth and allow adding things such as CPUs, RAM, disk drives and other things
• Rack mounts - servers should be rack mountable to save space and improve interconnectivity
• No side access needs - servers should be constructed so that the you only need front and back access for making use of servers

Remember, overall performance of a server isn't determined purely by the speed of its CPU.  You also need to consider a large variety of things such as hard drive access speed, graphics processing and display and input/output.

It is also important to go with reputable and appropriately priced vendors.  You should also expect vendors to help you do some of the configuring and fair amount of the service on your servers.

There are a lot of advantages to buying servers from a very few vendors.  Such environments are usually easier to maintain and interoperability and ease of configuration should be improved.  Buying from more than one vendor can improve negotiating power during sales.

Maintenance Contracts.  When buying servers it is also very important to consider how repairs will be performed.  Vendors have a variety of maintenance contracts with features associated with on-site service, response times, and warranties.

Some typical scenarios for picking the appropriate maintenance contracts are

• host with low/medium importance - service contracts can be for next or two day repairs.  It may also be the case that no special contract is needed because default options are sufficient.
• large groups of similar hosts - sometimes a site has many of the same type of machine.  In this instance it may be worthwhile to purchase a spare parts kit so that repairs can be performed in-house.  In this case, the hosts may now require a lower cost maintenance contract.  You may also get some economies of scale.
• controlled model selection - you might standardize the time horizon associated with purchasing particular servers.  This can also help in terms of maintenance and spare parts purchases.
• critical host - this is likely to require a service contract with a same day
• large variety of models from same vendor - in this case you may opt for a contract that includes an on-site technician supplied by the vendor.
• highly critical host - some vendors offer an option that includes an on-site technician and a duplicate machine ready to be swapped.  This is often as expensive as making use of a redundant server

There is a trade-off between how much you do your own service while stocking your own spares and how much your rely on a maintenance contract.  Sometimes it is easier to do your own troubleshooting.  But it is also very difficult to make sure your own staff is up to date on all the possible problems that can happen with your current hardware.

Sometimes a sys admin discovers that an important or maybe even critical host is not on a maintenance contract.  The following are three likely useful ways to prevent hosts from being missed on a contract.

• Have a good inventory system to track your acquisitions and cross reference it with your service contracts.
• Have the person responsible for making purchases be responsible for ensuring the appropriate purchases are on maintenance contracts.
• Make sure you are aware of the warranty duration and whether you need to add your purchase to a service contract when it expires.  The sys admin should see if the vendor can place the purchase on a service contract when it is purchased but show a zero dollar amount for the duration of the sales warranty.

Several other issues likely to be of significant importance are

• Data backups - which we will get into in much greater depth in a couple future chapters.
  • Clients are seldom backed up.  Hopefully, your automated or intelligent installations and updates take care of the major aspects of restoring a client.  Even with this, users are likely to have accumulated their own data and information on their machines.  This can be backed up in a variety of ways.
  • Servers need to be backed up quite frequently.  They are much more likely to contain mission critical information.
• Data Centers - which we will get into in much greater depth in a couple future chapters.
  • It is crucial to have things like fire and water protection for your servers as well as other important infrastructure items such as air conditioning.  You also likely to need power surge protection and alternative sources if power if there are power failures.
• Operating System
  • The OS doesn't need to be as uniform from server to server as it is likely to on desktops.
• Remote Access Administration
  • Servers need to be maintained remotely as much as reasonable.  You may need to be in physical contact with the machine in some instances.  But in most situations you would rather manage them from one location.
  • If configured correctly a single console server can be used in place of many dedicated terminals for each server.
  • Remote access capabilities for a server should be a major consideration during purchase decisions.
• Mirrored Root Disks
  • RAID - Redundant Array of Independent Disks can be used to mirror the main system disk and possibly other disks.
• Server Appliances
  • These are made to enhance performance for more specifically oriented servers such as for e-mail, web, DNS and so on.
  • They often have important specialized features.
• Redundant Power Supplies
  • While hard drives are the most likely failure point in servers, power supplies are the next.
  • Each power supply should have a separate cord
  • It is also fairly likely that you will want to have a UPS - uninterruptible power supply
• Full and n+1 Redundancy
  • It is often the case that you want to improve the reliability of critical systems.  Providing some redundancy can be crucial for this.
  • If there is some redundancy it is also likely to be possible to configure things so that there is some load sharing.
• Hot Swappable Components
  • Hot swap refers to the capability to remove and replace components while the system is running.
  • It is important to determine what components can be purchased and configured as hot swappable when purchasing servers.
• Separate Networks for Administrative Function
  • It is often the case that it is important to separate important administrative functions onto a separate network for a variety of reasons.
  • Information such as payroll needs to be almost entirely inaccessible to hackers and other gate crashers.
  • Certain types of administrative decision making or competitively advantageous information is also likely to need such additional efforts.