Easy-to-use performance tools with a consistent user interface across HP operating systems

Hewlett-Packard Journal, June, 1991 by Rex A. Backman

While each operating system has a measurement interface, the sets of counters differ. This difference presented a challenge for the HP GlancePlus development team because we had to develop the appropriate measurement interface for each platform and still present to the end user similar metrics regardless of the platform. The challenge was overcome by mapping the disparate measurement interface counters to a common set of performance metrics that all models shared. While some metrics are unique to each platform, these are in the minority. The majority of the metrics did map to common data items for each version of HP GlancePlus.

Another source of data is nonperformance-related items residing in certain operating system structures such as process-specific information not contained in the platform's measurement interface. Examples include process wait reasons, current process priority, and file system information related to the process. This information is used to provide discrete data on process or program related items. While not part of the measurement interface, these data items provide important information that can be used to aid in performance management. Maturity of the platforms dictated how this data was to be retrieved; MPE provides direct routines while HP-UX data is retrieved using a proprietary library of routines. These secondary data sources coupled with the platform-specific measurement interfaces provide each HP GlancePlus model with all required data items for computing performance metrics.

Binding the measurement interface as the primary data source with the secondary kernel data structures provides a simple and streamlined approach to the data gathering required for performance metrics. tying these data structures with relatively straightforward timing algorithms allowed the design team to meet the goal of creating easy-to-use performance diagnostic tools.

Processing Logic

A look at the internal logic of the HP GlancePlus programs shows a simple and efficient design. By having a core technology available, each model was able to leverage code from previously written software modules. The core section is an area of software procedures and programs that existed for other HP software engineering tools or products. The functionality provided by these tools is used to collect data from the measurement interface and kernel data structures for systemwide global and process-specific performance metrics. Some of the software modules that were reused had been written for the HP LaserRX family of performance products, and with very slight modifications they fit into the HP GlancePlus product easily.

Performance data at the global level includes systemwide information that can provide some insight into systemwide bottleneck conditions. Overall disk throughput, memory manager use, and global file open rates are examples of global metrics. Process-specific data includes data items that are associated with a specific process for the current interval. This data provides a mechanism by which to view a process in detail and compare it to the bigger picture of global performance data. As shown in Fig. 6, the modules responsible for collecting and summarizing global and process-specific data are in a refresh loop in which they are continually sampling, processing, and sending the data to formatters for display.