لطفا منتظر باشید ...
Analyzing High-Availability Requirements
After determining that you have mission-critical applications or services that need to be highly available, analyze your business needs to determine if clustering is the best solution. You must then evaluate your applications for compatibility with clustering. Figure 7.2 illustrates the process for determining if server clusters are appropriate for your needs.
Figure 7.2: Analyzing High-Availability Requirements
Analyzing High-Availability Business Needs
This chapter assumes that you have already decided to make one or more applications or services highly available, based on your organization's business needs. You need high availability if users require regular access to data or if your organization cannot tolerate downtime from an application or service for more than a minimal amount of time. Other common business motivations for high availability include contractual obligations, concern about losing sales during the time it takes to rebuild a server, and the potential for lost employee productivity when data is inaccessible.If your business or organization requires high availability, answer the following questions before you begin designing a server cluster:
How many users depend on a particular application?
Are your users located at more than one physical site?
What is the minimum acceptable downtime for a particular application?
How many failures does your cluster need to tolerate? One node failure? Two nodes or more?
How much hardware (particularly cluster nodes) can be dedicated to the server cluster?
Additionally, if you have more than one application that needs to be made highly available, do some applications take priority over other applications that will be running on the server cluster? If the answer is yes, consider prioritizing the applications. For example, if more than one node fails in your server cluster, and the remaining hardware is not sufficient to take on every failed node's workload, which applications and services must continue working?Understanding the business requirements that must be handled by your server cluster is important, particularly later in the server cluster design process when you design failover policies. For more information about these requirements, see "Determining How to Deploy Applications" later in this chapter.If a server cluster is the most appropriate high-availability solution for your business or organization, the next step is to make sure your application can run on a server cluster.
Determining Application Compatibility for Server Clusters
Not all applications can be adapted to run on a server cluster. If your application meets all the following conditions, it will likely work in a server cluster:
Your application uses a protocol based on Internet Protocol (IP).
Your application can specify where application data is stored.
Your application attempts to reconnect or recover from network failures.
IP-based protocols include Transmission Control Protocol (TCP), User Datagram Protocol (UDP), Distributed Component Object Model (DCOM), named pipes, and remote procedure call (RPC) over TCP/IP. NetBEUI and Internetwork Packet Exchange (IPX) protocols do not work on a server cluster.
| Tip | For information about pretesting to determine if your application is cluster ready, or for information about whether your application supports failover, whether clients can survive failure and subsequent restart of your application, and if clients can survive a failure without a restart of your application, see "Choosing applications to run on a server cluster" in Help and Support Center for Windows Server 2003. |
Additionally, some applications do not react to cluster events. An application must be cluster-aware (must communicate with the Cluster application programming interface) to receive status and notification information from the server cluster. The Cluster application programming interface (API) is a collection of functions that are implemented by the cluster software and used to manage the cluster, cluster objects, and the cluster database.Most older applications are not cluster-aware, but they can still run in a server cluster and fail over. These applications respond to only the most basic methods of failure detection and shutdown, and they are not capable of the more advanced initialization and cleanup tasks performed by applications that can react to cluster events. For this reason, these applications might not be as highly available in a server cluster as are applications that can interact with the cluster.For more information about applications that can run on a cluster, see the Software Development Kit (SDK) link on the Web Resources page at http://www.microsoft.com/windows/reskits/webresources and search for the help topic "Server Cluster Application Types."Before designing your server cluster, contact the software vendor or consult the documentation to see if your application can run on a cluster. This information can affect how you install your application in a server cluster, and might also determine the best way to upgrade an application once it is installed in a server cluster.
