A baseline defines the typical activity of your network servers. Keeping a baseline document of activity on a server lets you determine when the activity is atypical. To create a baseline activity, you should gather statistical information when the server is functioning typically.
This section contains the following information to help you plan your server management strategy:
For server statistics such as CPU utilization, you should create a trend graph that plots information over a period of time. Statistics sampling that gathers data over a short period of time can be misleading. If you modify the server's configuration, it is useful to create another baseline against which you can compare future activity.
There are two ways to create baseline documents. The first is to create them manually by printing the various trend graphs for which you want to maintain baselines. The other way is to use the server management health reports as your baseline documents. For more information on creating and generating health reports, see Managing the Server Management Health Reports. In either case, the data gathered can be exported into programs, such as spreadsheets, for further analysis and to maintain records over time.
The following sections will help you plan and use the baseline document:
You should set alarm thresholds for statistics on servers monitored by the SNMP agent software, so that if the threshold is exceeded, you are notified at Novell ConsoleOne. Setting alarm threshold values for statistics on a server eliminates the need for you to constantly monitor polled server statistics for problems.
Server Management components provide default values for thresholds set on server statistics; rising and falling statistics generate an alarm when a threshold is surpassed.
By comparing current server performance statistics against the performance recorded in your baseline document, you can determine how performance is affected by server configuration changes. This comparison also helps you plan for growth and justify upgrades and expansion. You can view graphs of real-time trends and historical trends over hourly, daily, weekly, monthly, and yearly periods.
You should include the following key characteristics in each server baseline document:
The CPU Utilization statistic indicates how busy the microprocessor is. High CPU utilization can cause slow network response time. Utilization is likely to be higher at some times during the day (for example, when users log in to the network in the morning, or access e-mail), week, or month. Tracking CPU utilization helps you track the load on the server processor at peak and low times. This information helps you determine the effect of current system and application processor demands and analyze the impact on performance.
Virtually all processes are handled through server cache, a block of server memory (RAM) in which files are temporarily stored. Cache buffers greatly increase server performance and enable workstations to access data quicker because reading from and writing to memory is much faster than reading from or writing to disk. The optimum cache buffer is 65% to 75% of total server memory (more does not hinder performance). Low cache buffers can cause slow server performance and abend. Service degrades noticeably at 45% of total server memory.
By tracking data about file reads and writes in your baseline, you might be able to determine whether a bottleneck is caused by the disk I/O channel. For example, if an increasing number of "server busy" packets are sent to users and there is also an increase in the file read and write number, the cause of the bottleneck might be a slow disk I/O channel or bad disk adapter driver.
Tracking volume utilization is primarily for capacity planning. By tracking the volume space used over time, you can accurately predict when you must purchase additional storage. Tracking volume utilization can also help you prevent the server from running out of disk space.
By including information about running software in your baseline, it is easier to spot a problem application when comparing software on different servers. It is useful to also include the memory each application uses. Then, if the server is running short of memory, you can quickly see which applications are using the most memory.