Ensuring the Availability of Applications and Services

Identifying High-Availability Needs for Applications and Services

To identify specific high-availability needs for mission-critical applications and services in your organization, determine the following:

Basic characteristics of the application or service, such as:
- Software used.
- Special hardware requirements. (For more information, see "Determining Hardware Compatibility for Advanced Features" later in this chapter.)
- Volume of data.
- Number of users.
- Hours of operation.
Expected changes in size and performance requirements, such as:
- Seasonal or other planned peak loads.
- Expected rate of increase in users.
- Expected rate of increase in data.
Hardware requirements at initial deployment, during peak load times, and within the planning horizon of the project. Use application and service characteristics and change projections to make your estimation.
Data backup and disaster recovery plans for the application, service, and operating system.
Maximum outage tolerance (that is, the maximum amount of time that the system or users can tolerate the application or service being unavailable ). This is important, because eliminating all possible outages can be expensive. For many applications and services, occasional brief outages might be an acceptable alternative to larger investments in high availability.
Impact of downtime. Perform a qualitative assessment of the way that downtime of an application or service would affect your organization. Examples include lost sales, lost productivity, and decline in customer satisfaction.
Measurable cost of downtime. Perform a quantitative estimate of the cost per application or service outage that exceeds the specified maximum that your organization can tolerate.
Identify all potential single points of failure in the planned configuration.
Identify your staffing requirements.
Current availability. If you do not have the availability numbers, begin collecting the data right away. Starting with Windows NT 4.0 SP4, you can easily calculate the system uptime using the Microsoft Windows NT 4.0 Server Resource Kit tool, Uptime.exe. Uptime.exe uses events stored in the Event Log to calculate these numbers. You must enable the Event Log in order for this to take place. Uptime.exe is also available in the Microsoft Windows 2000 Server Resource Kit.

A single point of failure is any component in your environment that would block data or applications if it failed.

Table 18.1 lists common points of failure in a server environment and describes whether you can protect the point of failure by using a Microsoft clustering solution or by using a third-party solution.

Table 18.1 Common Points of Failure

Failure Point	Clustering Solution	Other Solutions
Network hub	N/A	Redundant networks
Network router	N/A	OSPF
Power outage	N/A	- Uninterruptible power supply (UPS) - Generator - Multiple power grids
Server connection	Failover	N/A
Disk	Fail over	Hardware or software RAID, to ensure against the loss of specific data on a specific computer and to provide for uninterrupted service
Other server hardware such as CPU or memory	Failover	Spare components such as motherboards and small computer system interface (SCSI) controllers (any spare components need to exactly match the original components, including network and SCSI components).
Server software such as the operating system or specific applications	Failover	N/A
Wide area network (WAN) links such as routers and dedicated lines	N/A	Redundant links that provide secondary access to remote connections
Dial-up connection	N/A	Multiple modems and Routing and Remote Access

Determining Hardware Compatibility for Advanced Features

Check to be sure the computer systems and adapters that are presently installed or are planned for purchase are listed on the Microsoft Hardware Compatibility List (HCL). To determine if the hardware is listed and supported by HCL, see the Microsoft Windows Hardware Compatibility List link on the Web Resources page at http://windows.microsoft.com/windows2000/reskit/webresources.

If you plan to deploy computer systems with advanced capabilities, such as clustering or large amounts of memory, there are additional requirements that you must meet.

For example, if you have Advanced Server computer systems with more than 4 GB of RAM, you must modify the Boot.ini PAE switch to enable PAE memory use. You can make this modification after you have verified that all components are supported. If the components are not supported, the computer system must be brought into compliance to avoid potential problems later.

If the system is in production and you add new adapters and drivers, it is strongly recommended that you back up the system completely before making any modifications.

Note

Changes to the registry and some modifications to the Boot.ini file that could cause the computer system to function incorrectly or be unstable, might be avoided when you start the computer by using the F8 key, which bypasses many switches and drivers. This allows you to make changes to the Boot.ini file or other areas as needed to restore functionality

Determining Your Clustering Requirements

After you have determined specific high availability needs for mission-critical applications and services, identified potential single points of failure, and determined that your hardware is compatible with Windows 2000, you must determine which type of clustering technology best fits the needs of your organization. Before you plan your cluster, review the requirements to determine the appropriate cluster type.