Number of CPUs
Most SQL Server operations take advantage of multiple processors, but you have to determine how much spillover goes to the second or subsequent processors. For most workloads, Task Manager will reveal that CPU activity occurs on the first processor with little on the second processor and even less on the third and fourth. If at all possible, study a representative workload to see how many processors you must have to satisfy needs. You can save on the per-processor licensing costs by using fewer processors, and then add them using the affinity switch as your workload increases.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
In general, most complex SQL Server operations (i.e., joins, aggregate operations, etc.) are processor intensive, and performance will improve on higher CPU systems. Study the processor queue length to determine if CPU is a bottleneck. If the queue length will consistently be above 2, you should consider using additional or faster processors.
L2/L3 cache is temporary processor storage used by the microprocessor. For processor-intensive workloads, select the largest L2/L3 cache available. Currently, the largest cache on Xeon processors is 1 MB. The new Itanium processors have an L1, L2 and L3 cache -- the largest L3 cache is 9 MB.
Spec your SQL Server hardware needs
Invest in good application design
Understand your workload
Know your memory support limitations
Choose a reliable hardware brand
Take advantage of 64-bit
Take advantage of storage area networks
Properly configure your RAID arrays
Use separate disk controllers
Choose and optimize your disks wisely
Optimize CPU activity and speed
|ABOUT THE AUTHOR:
| Hilary Cotter
Hilary Cotter has been involved in IT for more than 20 years as a Web and database consultant. Microsoft first awarded Cotter the Microsoft SQL Server MVP award in 2001. Cotter received his bachelor of applied science degree in mechanical engineering from the University of Toronto and studied economics at the University of Calgary and computer science at UC Berkeley. He is the author of a book on SQL Server transactional replication and is currently working on books on merge replication and Microsoft search technologies.
Copyright 2006 TechTarget