Performance Monitoring Overhead

There are several techniques you can use to ensure that performance-monitoring overhead is kept to a minimum on any database server you are monitoring.

The System Monitor application can be demanding on resources. It may be better for you to use logs than to display graphs. You can then import the data into report programs and databases for later analysis. Save the logs to storage that is not being monitored, or to a hard disk that is not the object of some form of analysis. Also make sure the logs are not growing too big. If you are restricted from setting a quota or an alert on the disk space, make sure to keep your eye on the disks.

Do not overdo it on the number of counters you are using at the same time. Some counters are costly on performance and can actually increase overhead, which will be counterproductive. It is also tough to monitor many things at once. Factor in what each counter consumes in overhead. This information is available in the Windows Server 2000 support CDs.

Tight collection intervals can also be costly, and you can over-monitor if you get too eager. Microsoft recommends that you leave a ten-minute interval between data collection periods.

Keep monitoring during peak usage periods to obtain the best assessment of resource usage. And at the same time, take care not to impact available resources. Naturally it makes no sense to monitor a system that is idle, and that is something you should only do if you need to know what the server looks like when it is not under load.

You should also monitoring remotely. Remote monitoring allows for a centralized data collection. You can also collect data from several servers and save the data to the local machine or a local database server. However, network bandwidth increases as you collect more data and you collect it more often. So consider keeping the number of servers in a monitored group to no more than between 10 and 15. To increase network bandwidth, consider saving the remote data to log files on the remote servers and then either copying the data to the local computer or database or viewing it remotely.

Finally, I guess there is also a case for being too cautious. As I said earlier, monitoring is in itself resource intensive. As long as you frequently gather data and analyze, it you’ll be okay.