Veritas NetBackup™ Appliance Capacity Planning and Performance Tuning Guide
- About this Guide
- Section I. Capacity planning
- Section II. Best Practices
- Section III. Performance tuning
- Section IV. Quick reference to Capacity planning and Performance tuning
About diagnosing performance issues
When encountering a performance problem, the first step is to try reproducing the problem. If the problem is reproducible, the next step is to collect the four major system resource usage statistics while the problem is being reproduced. The four major system resources are:
CPU
Memory
I/O
Network
The most common utilities that are used to collect resource usage statistics are:
vmstat
iostat
ifstat
With these three commands running concurrently, you can get a snapshot of the usage pattern of these four critical system resources. Examining the usage pattern of these resources could reveal the entry point to further performance issues. These utilities must be run concurrently with the same reporting interval. You can run these three utilities in the following manner.
Open three terminal windows and run each command in a separate window. In the NetBackup Appliance Shell Menu, go to menu to run these commands. Enter a password when prompted.
Note:
These commands must be run concurrently with the same reporting interval.
Run the following commands:
vmstat 5 60 | tee /tmp/vmstat.5
iostat - kxdt 5 60 | tee /tmp/iostat-kxdt.5
ifstat 5 60 | tee /tmp/ifstat.5
When you run these commands, a copy of the data is saved in the /tmp directory.
Note:
Ensure that there is enough space to save this data. Running out of space can disrupt normal operations.
Note that the number 5 after the command is the interval in seconds, while the second number 60 specifies how many reports to collect. You should not use the first report, because it gives the average statistics since the system was last started. You can change the interval (seconds) and the number of reports. In general, an interval of 5, 10, or 15 seconds is recommended for online monitoring. For a longer period of monitoring, you can use 30 seconds or 60 seconds to reduce the disk space needed to store the statistic files. Since the reports are the average over the interval, the best practice is to always collect the statistics using the same interval. This will help you develop a sense of how system resources are being consumed when the system is running normal. With this knowledge it will be easier to spot any change in resource usage and help find the root cause of a performance problem.
The following sections show how to detect potential performance bottlenecks from the output of the above three commands, and also the actions you can take to mitigate a performance problem. See the man page of these commands for detailed information about the output fields.