How much space to allocate

Data written to the persistent data store is organized by tables (attribute groups), groups, and data sets. Each table is assigned to a group; each group can have one or more tables and one or more data sets assigned to it. The size of the persistent data store (spread across the data sets), the groups that are selected, and the activity of the system or subsystem monitored determines how much data is available. By default, one group and six data sets are allocated. When you configure the persistent data store, the persistent data store processing computes how much space is required for the group data store files and how much additional required space is required for overhead information. (Overhead information includes the product dictionary, table records, index records, and spare room for buffers that must be reserved for when the data set is full.) However, you might have to adjust the space allocated.

Initially, you should accept the default number of cylinders and groups provided by each product (Kpp_PD_CYL=, Kpp_PD_GRP=). Eventually, you can determine the correct amount of space by observing how often the maintenance procedures are running and adjusting space according. To help you make more specific calculations, the product documentation for the monitoring agents provides estimated space requirements or information about attribute tables. You might want to make your own calculations, based on site-specific factors: what types of monitoring agents are running, what resources are being monitored, how many resources, and so on. You can also take a trial-and-error approach. If you are not seeing as much data in your reports as you would like, you can override the defaults.

Increasing the group count (number of data sets used to provide the total number of cylinders of persistent data store space) is a way to reduce the total space required, by reducing the percentage of the total space required for the in-use and empty data sets. Increasing the group count to 6 or 8 dramatically reduces the total space required. (The maximum group count value is 36.) Another benefit of increasing the group count is that roll-off to the warehouse as well as short-term history queries perform better when the individual data sets are smaller. The overhead for more frequent data set switches caused by the increased group count (and presumably by executions of the maintenance procedures) is more than offset by the gains in efficiency.

When historical data is collected in generic history data sets at the monitoring server, the same considerations apply, but the data sets must be sized to contain the data from all of the agents that write to them.

For more information on changing the size of the persistent data store, see Resizing persistent data store files and How to: Change PDS file count and file size.