Rick Heiges

The old Hardware Compatibility List which was renamed the "Windows Catalog" is no more.  When creating a Clustered Solution, you need to have a system that is Certified for clustering or there will be no official support.  This has been a point of pain in the past as organizations purchased a clustered solution and then one of the nodes has issues after a year or two.  Getting an exact replica of that node is often impossible and impractical.  Try explaining to the big wigs that you need to purchase a new system that is 2 years old - not very compelling.

In Windows 2008 and SQL 2008, there is no need for a certified solution.  There is a program called clusprep.exe that you run on the nodes of the proposed cluster to test/prepare the nodes.  If your system passes, it can run on a cluster. 

One drawback of this is that when failing over to a node of lesser "girth", the user experience may suffer.  Also, with this change, the base of clustering is changing which means that your upgrade to the 2008 flavor will take some more planning.

If you attended the PASS Community Summit in Denver, you can view the streaming session online on Clustering.  It was very educational.

I started the Triad SQL Server User Group with several other folks back in 2002.  I led the chapter for 2 years before turning the reigns over to a Board of Directors.  I have only been to one or two meetings since the turnover mainly because I have been out of town when they have meetings.  But last night, I was in town and the Holiday Social was just 2 miles away from my house at a local drinking/dining establishment. 

It was good to see the Board of Directors and meet some of the regulars who now attend. I was so happy to see that the group has really grown over the past few years in its sophistication, marketing, and breadth of topics.  I wanted to acknowledge and praise the team that is now the Board of Directors - Miguel Cebollero, Michael Wharton, and Karen Lennon.

All of us have heard that if you just add memory to your system, it will work faster.  This is generally accepted and it makes alot of sense, but what is the math behind it?

About 2 years ago, I heard the presenter of a webcast that I was viewing put these two elements in a perspective that I relay on to others when speaking about memory.  Basically, Disk Access is measured in milliseconds; RAM Access is measured in nanoseconds.  There are 1 million nanoseconds in 1 millisecond.  In other words the magnitude of access time is on the order of 1 million; that's pretty big.  In real human time, I can't distinguish between a nanosecond and a millisecond, but there is a HUGE difference. 

Just about everyone can look at their watch or a clock and know how long a second takes.  If you sit and stare at that timepiece until 1 million seconds tick by, you will stare at it for 11.57 days!

(1 million seconds / (60 seconds) / (60 minutes) / (24 hours)) = 11.57 Days

As you can see, there is a HUGE difference in the access speeds between IO and RAM.  We all knew this inherently, but sometimes an example can really help - especially when budgeting season is imminent.