- What type/brand of hardware I have to purchase?
- Is it feasible to place one server in the North and one server in South for clustering?
- What are key considerations for clustering?
- Is there another way other than clustering to keep system online?
Clustering is not a one-size-fits-all solution for high-performance computing; the main reason to create a cluster is to provide redundancy, so that if one computer fails or must be taken offline, the other computer can automatically step in and resume operations. If you require high availability from your servers, this is one of the ways to do it. The best way to get a clustered system is as a pre-built system from a Microsoft-approved vendor (assuming you are using Windows and not Unix), such as HP or Fujitsu. Geographically, the two systems do not have to be in the same room; they can be separated by a network link, although for the best results it should not be a WAN link. Other than clustering, another method for building a high-availability system is fault tolerance. Mirrored disks, hot-pluggable memory and processors, and other hardware-level features make this possible now, but I believe only Windows Server 2003 now supports these features innately.
This was first published in June 2003