Managing Collection of Metrics
webMethods API Gateway collects performance data (for example, average response time, total request count, fault count) for the virtual services that it hosts. It publishes this data to the destinations (as an example, CentraSite, API Portal, Elasticsearch, and EDA/Database) at regular intervals. When you install and configure the API Gateway, you must specify whether you want it to collect performance data and, if so, how often you want it to publish the data to CentraSite.
We recommend that you always enable the collection of performance data on your API Gateway. A publication interval of 15 minutes is appropriate for most environments. However, if API Gateway handles a very high volume of traffic, consider increasing this interval to 30 or 60 minutes. CentraSite stores the performance data that it receives from the API Gateway in the performance log. You can look at the performance information for a particular virtual service by viewing its Runtime Metrics profile in CentraSite Business UI.
The performance data that CentraSite collects from API Gateway can cause the log to grow quite rapidly. When the log grows very large, queries to the log can significantly affect CentraSite's performance. To prevent this from happening, we suggest that you routinely purge old entries from the log.
CentraSite provides a log-purging utility that you can use to automatically purge the log on a scheduled basis. We suggest that you use this utility to keep no more than one month of performance data in the log (adjust this recommendation as necessary to accommodate your particular needs). When you configure the log-purging utility, you can specify whether you want to delete the purged log entries or export them to an archive file (in case you want to retain them for future reference).
Note:
The performance metrics that API Gateway collects enable service consumers (and potential service consumers) to determine whether a virtual service is performing at a required level. However, API Gateway does not collect data at the granularity that a network administrator would need in order to analyze performance problems (for example, to determine why the response time for a particular virtual service drops at a particular time of day).