The Microsoft Stack: Windows Server 2008 Clustering Designed to Overcome Usability Barrier

By Patrick Hynds

August 1, 2008 —  (Page 1 of 3)

Reliability is important to all organizations, but in some cases it is life and death. Earlier this year I helped a health-care organization recover from an outage that had an entire hospital working without crucial data they would rather not lose access to while treating patients.

The offending system was a Windows 2003 Server cluster hosting SQL Server, as well as other services. I was called in not to fill a gap in technical depth, but because Microsoft clusters are a bit of a black art if you don’t deal with them regularly. We were able to recover the offending node and restore the corrupt shared registry in a very short time, but it highlighted the real barrier to why more organizations are not using Windows clusters more widely, and that is usability.

Cluster servers started with a product years ago that was code named Wolf Pack, a very World War II-style code name, and has evolved to be quite powerful but not terribly easy to use. The good news is that ease of use is a major goal with the new version of clustering soon to be available in Windows Server 2008.

Before I dive into the promised improvements, let’s review the concepts behind Windows Clustering, starting with the quorum drive. To do this, let's think of two people working to ensure a critically ill patient receives a high level of care for 24 hours. The key element to ensure the continuity of care is that both caregivers share a common knowledge of the situation at any given moment and also that they know who is actually supposed to be in charge at any given time, to avoid such potentially deadly events as giving overdoses of medicines.

The quorum drive provides the storage location that both nodes share (by turn), so that when one fails, the other can pick up where things left off. There are also shared resources that fail back and forth as the active node changes. For this reason, any stateful objects like files or logs are stored in a place that the active node can assert control over and pick up where the failed node left off.

Related Search Term(s): clustering, Windows Server, Microsoft

Share this link: http://www.sysmannews.com/link/32626