Job Management Partner 1/Performance Management - Agent Option for Platform Description, User's Guide and Reference
This subsection provides an overview of performance monitoring.
- Organization of this subsection
- (1) Objective of performance monitoring
- (2) Most important system resources in need of performance monitoring
- (3) Determining baselines
(1) Objective of performance monitoring
The monitoring of performance is an important task in the creation and management of a system environment.
You can achieve the following major performance monitoring objectives by using PFM - Agent for Platform:
- Determine the causes of bottlenecks (by analyzing performance data).
- Understand load characteristics and their impact on the corresponding system resources (by analyzing performance data trends).
- Verify whether the system is operating correctly (by monitoring it).
Certain types of bottlenecks can affect the operating performance of the entire system, such as bottlenecks produced by the following causes:
- Insufficient memory
- Monopolization of specific resources by programs
- A damaged subsystem
- An incorrect subsystem configuration (for example, 10 Mbps is set for a 100-Mbps NIC)
- Unbalanced subsystem loads
By monitoring performance under different conditions (such as different numbers of users connected concurrently) and by monitoring performance continuously, you can check the load characteristics and the impact on the corresponding system resources in the system environment. The benefits of such monitoring are as follows:
- Changing and adjusting the system configuration are simplified.
- System resource upgrades can be planned.
Checking whether a production system is operating correctly is extremely important. You can check whether the system environment is operating correctly by using the following types of monitoring together with performance monitoring:
- Monitoring of the operation of processes that provide system functionality
- Monitoring of the operation of services that are prerequisite for the system
- Monitoring the system for invalid processes
Using PFM - Agent for Platform as described above to correctly monitor performance is vital for stable system operation.
This section explains how you can use PFM - Agent for Platform to monitor performance to ensure stable system operation.
This section uses examples to explain the various methods of performance monitoring. Note, however, that the thresholds used in the examples are not applicable in all situations. Accordingly, actual thresholds must be determined only after baselines are selected. In addition, other settings will be dependent on the operating requirements of the system.
(2) Most important system resources in need of performance monitoring
The performance monitoring performed by PFM - Agent for Platform treats the following system resources as being the most important:
- Processor
- Memory
- Disks
- Network
- Processes
- Services
In addition, PFM - Agent for Platform uses the following information, which is useful for detecting problems with the OS and abnormal processes:
- Event logs
PFM - Agent for Platform provides a monitoring template containing these important items. This section focuses on explaining how monitoring templates are used to perform monitoring. This section also provides related information for advanced monitoring.
(3) Determining baselines
Determining baselines is a task in which system values unlikely to affect normal system operation are determined from the performance measurement results.
PFM products use baseline values as thresholds for monitoring system operation. The determination of baselines is therefore an important task for successfully monitoring performance.
Note the following recommendations when you determine baselines:
- Measure system performance at peak load (for example, when a high-load test is run for the operating environment).
- Thresholds depend greatly on the system configuration. Therefore, performance should be measured again to redetermine the baselines if system resources and the operating environment have been changed.
All Rights Reserved. Copyright (C) 2009, Hitachi, Ltd.