You are the administrator of a Microsoft Windows NT Server 4.0 computer. The server is the PDC in your domain and also runs SQL Server 2000. The server has four processors and 1 GB of RAM.Your network consists of 500 client computers that are running Microsoft Windows Millennium Edition. Users report that when they log on in the morning their computers respond slowly and occasionally return error messages.You use System Monitor to monitor processor performance and RAM utilization. You discover that the % Total Time and % User Time counters average 90 percent and the % Privileged Time counter averages 20 percent. The Available Mbytes counter averages 800 MB. You need to improve server performance for user logons. What should you do?

The system monitor clearly indicates that the processors are overloaded. We need to decide which processors we should use for SQL Server. Processor 0 is the default CPU for the I/O subsystem. Network Interface Cards (NIC) are assigned to the remaining CPUs, starting from the highest-numbered CPU. The NIC would get processor 3.
SQL Server would be using all four processors by default.

The Windows NT/2000/XP operating system use processor 0. In order to avoid the logon
problems we should not let SQL Server to use this processor. Instead SQL Server should be configured to use processor 1, 2, and 3.

Note 1: The Affinity mask is used to exclude a processor on a multiprocessor computer from processing SQL Server 2000 threads. Default is equal distribution of SQL Server 2000 processes across all processors.

Note 2: Monitoring the Processor and System object counters provides information about the CPU utilization and helps in determining whether or not a bottleneck exists. The %Total Processor Time is used on multiple CPU Severs to gauge the average activity of the processors and shows the percentage of elapsed time that a processor is busy executing a nonidle thread. Values around 100 percent on a Server computer that processes many client requests indicate that processes are queuing up, waiting for processor time, and causing a bottleneck. Such a sustained high level of processor usage is unacceptable for a server.