You are on page 1of 3

Hardware Performance Counters

Commerce Server 2007


For the latest version of Commerce Server 2007 Help, see the Microsoft Web site.
The following table describes the hardware performance counters you should use to diagnose
and monitor the performance of your Commerce Server computer. You can find these counters
in Microsoft Windows Server System Monitor. You should monitor each server in your data
center with performance counters and you should view the results collectively. You can use the
counters in the System and Memory objects to monitor capacity.
Performance Counter
Description
object name
Bytes
received per If a network card approaches its maximum capacity, you should
Network second add another card. If this value approaches the capacity of the
Segment
Bytes sent network, a higher bandwidth network might be necessary.
per second
Inetinfo Monitor this for memory leaks or size approaching maximum
Instance (in the Virtual bytes available RAM. This counter measures the virtual memory
Process object) allocations.
This should be well within the hardware capacity you have
Working set
available, which means less than the physical RAM.
Available
Memory This should be more than 4 MB.
bytes
This should be less than one page per second. If the system is
Pages per
actually out of memory, this is the biggest indicator of the
second
problem.
Disk
Writes/secon
d Combined, these two counters should be well under the maximum
Physical Disk
capacity for the disk device.
Disk
Reads/second
The percentage of elapsed time that the selected disk drive is busy
% Disk Time
servicing read or write requests.
IO
SQL Server transactions/s Indicates how much activity the SQL Server actually performs.
ec
Compiles/sec Indicates the efficiency of the queries that SQL Server is running.
Recompiles/ Reducing this number is very important to reducing the CPU load
sec on the SQL Server.

Context If this rate is more than 5,000 context switches per CPU, you
System
switches/sec should add another server to horizontally scale the system.
Peak utilization can be 100 percent, but you should not sustain
%Total
utilization at a level past the capacity of your processor. You can
Processor
scale all the server elements horizontally.
Total The number of active Transmission Control Protocol (TCP)
Web
connections connections to the server.
Get The number of HTTP requests that the server receives. This is the
requests/sec primary indicator of traffic to a Web site.
Non-
anonymous
Tracks the number of authenticated user requests at a site.
users per
second
Active Server Requests per The rate at which the Active Server Pages (ASP) are processing
Pages second requests. Includes both successful and failed requests.
Requests
The average number of requests running at a time during the
executing
sample period. Internet Information Services (IIS) 6.0 has many
hardware
threads that can simultaneously process requests.
counter
This should be very close to zero (less than 100 milliseconds)
Request wait because this is the period of time a request sits in the queue before
time it starts processing. You do not want users to experience long wait
times to process their checkout transactions.
Requests
There should not be a significant queue except at peak periods.
queued
Bytes Received Per Second Hardware Counter

Compare the bytes received per second counter to the total bandwidth of your network adapter
card to determine whether your network connection is creating a bottleneck. To allow for room
for spikes in traffic, you should usually use no more than 50 percent of capacity. If this number is
very close to the capacity of the connection, and processor and memory use are moderate, the
connection might be a problem.
Requests Executing Hardware Counter
If the number for the Requests executing counter is always greater than one (1), there could be a
problem in an ASP page or object that is causing a request to stop responding. If this number is
never greater than one, even during peak periods, you might have accidentally left ASP bugging
on in your site.

You might also like