|
|
So why is clustering significant? Imagine running a Web site on only one server. If that server went down, the whole Web site would stop working. A single server also has a limit on how many concurrent users it can handle before the load degrades performance so much that it becomes a real problem. Clustering has been developed to solve these two problems, which we call availability and scalability, respectively.
Availability means that a service, say an e-mail service, should be available for the users for as much time as possible. Imagine the anger among Hotmail users if the service went down for long periods of time, making it impossible for them to check their e-mail during this time. That would not be so good. To solve this, the administrators of the service must make sure that no single point of failure exists, because if it does and this point fails, the service also goes down. Often the solution is distributing the service on several servers to avoid this horror.
The second problem is scalability, which means that if the workload on a service increases, you must be able to add more power to handle the increase. Imagine the e-mail service again. What if your e-mail service provider got so many new users that the provider's servers could not handle this new load and crashed? That would not be a dream scenario at all. To avoid this, providers often add more CPUs or more computers to the cluster.
|
|