Every five minutes, the S Series Node issues an internal request for system-load statistics and saves the returned statistics, along with a timestamp, in the internal database. Once an hour, these minute statistics are rolled up into hourly statistics, and, a week later, the minute statistics are deleted. Once a day, the hourly statistics are rolled up into daily statistics, and, three weeks later, the hourly statistics are deleted. Once a month, the daily statistics are rolled up into monthly statistics, and, three months later, the daily statistics are deleted. Monthly statistics are kept for two years.
Each rollup consists of the average, minimum, and maximum for each type of statistic for each of server module 1, server module 2, and the S Series Node as a whole during the applicable period. For example, suppose the five-minute values for CPU load on the S Series Node in a given hour are 25%, 50%, 30%, 90%, 25%, and 50%. Based on these values, the rollup for the S Series Node for the hour would have a minimum value of 25%, a maximum value of 90%, and an average value of 45%.
You use the management API /metrics/systemLoad/history resource to request historical system-load information from an S Series Node. In the request, you specify the start and end times for the period of statistics you want. You can also specify the granularity (minute, hour, day, or month) of the statistics you want. If you don't specify the granularity, the S Series Node returns hourly statistics.
The response to a request for the /metrics/systemLoad/history resource can be quite long. You can use query parameters to limit the response in these ways:
- Omitting the S3 compatible API statistics
- Omitting the minimum and maximum values for rolled-up statistics
- Omitting the server-module-specific statistics
- Including only statistics you specifically request (for example, if you request only CPU and memory statistics, no statistics are returned for network bandwidth, I/O threads, or the S3 compatible API)
- Paging through the response by repeatedly issuing the request, each time specifying a limited number of results to include in the response
Omitting the S3 compatible API statistics and including only statistics you specify are mutually exclusive options.