This plot is a time series representation of the use of micro plots to illustrate the sizing decisions DB2’s memory manager is making for sizing multiple buffer pools (left series of plots, x-axis is time, y-axis is pages) and the resulting logical read rates (right series of plots, x-axis is time, y-axis is reads) issued against those pools. The source of this data is the db2diag log, a text file containing verbose, multi-line entries documenting these buffer pool statistics.

<log snip here>

While the log is a rich data source, it’s not easy to visually parse for interactions between these buffer pools. A simple script to parse this data into data tables capturing the pool sizes and read activity was used to pull the data from the log files. Once the data was extracted, another script was used to built the plots.

These scripts were part of a broader stack overlay that allowed for the dynamic evaluation of workload behavior from multiple data sources, including this db2diag data, application tier sar, iostat, and application latency data, and database tier sar, iostat data. The red bars illustrate the resizable and moveable sample intervals that are summarized by the counts next to the plots.

The use of simple scripts combined with time series plotting strategies can greatly simplify the summarization of large, complex data sets.