Here is a list of best practices which we strongly recommend to follow to get the best results, performance and best end-user experience.
Amount of indicators and attributes in one data set
Keeping reasonable amount of indicators and attributes inside a single data set may help the end users to keep their analytics in well organized form. From the performance point of the view, more objects need more HW resources and may lead to performance slow downs if the limit is reached. Recommended amount of attributes/indicators defined within one data set is 100. Generally having 100 objects inside one data set is anti-pattern, the same when having more than 10 views inside one report. This may lead to the performance issues. There must always be a way how to define the reports and data set in more efficient way - having more data sets and more reports to reduce complexity of single data and single report.
Maximum number of indexed attributes
Having more than 50 indexed attributes (indexed = intended to use for aggregations and filtering, string, date, time, geopoint) may have deep impact on the report calculation response times. We recommend to carefully distinguish between indexed and not indexed attributes on the import settings stage. Descriptive attributes like long texts or GeoJSON should use appropriate attribute types.
Number of views in each report
Best practices for formulas
Limits in tables in multi-domain cloud
When you are running multi-domain cloud with multiple instances, make sure that users do not define custom limits for attributes in tables which are used for exports. In case they are running many exports with large tables, performance issues might appear. This point is not related to single domain installations.