Thomas Kejser, Microsoft Program Management
Reviewers
Prem Mehra, Microsoft Program Management
Steve Howard, Microsoft Program Management
Paul S. Randal, SQLskills.com
Kun Chen, Microsoft Development
Gus Apostol, Microsoft Program Management
Sanjay Mishra, Microsoft Program Management
Alexei Khalyako, Microsoft Program Management
Diagnosing and Resolving SpinLock
Contention Issues
Historically, commodity Windows Server computers have utilized only one or two
microprocessor/CPU chips, and CPUs have been designed with only a single processor or “core”.
Increases in computer processing capacity have been achieved through the use of faster CPUs,
made possible largely through advancements in transistor density. Following “Moore‟s Law”,
transistor density or the number of transistors which can be placed on an integrated circuit have
consistently doubled every 2 years since the development of the first general purpose single chip
CPU in 1971. In recent years, the traditional approach of increasing computer processing
capacity with faster CPUs has been augmented by building computers with multiple CPUs. As of
this writing, the Intel Nehalem CPU architecture accommodates up to 8 cores per CPU, which
when used in an 8 socket system can then be doubled to 128 logical processors through the use
of hyper-threading technology. As the number of logical processors on x86 compatible computers
increases so too does the possibility that concurrency related issues may occur when logical
processors compete for resources. This guide describes how to identify and resolve particular
resource contention issues observed when running SQL Server 2008 and SQL Server 2008 R2
applications on high concurrency systems with some workloads.
In this section we will analyze the lessons learned by the SQLCAT team from diagnosing and
resolving spinlock contention issues, which are one class of concurrency issues observed in real
customer workloads on high scale systems.
In This Section
Symptoms and Causes of SQL Server Spinlock Contention
Diagnosing SQL Server Spinlock Contention
Options and Workarounds for Handling SQL Server Spinlock Contention
Appendix A: Automating the Process of Capturing Memory Dumps to Analyze Spinlock
Contention
Appendix B: Capturing Spinlocks Statistics Over a Specific Time Period
