Experience Optimization scaling recommendations
Follow these considerations and guidelines when scaling Experience Optimization and OpenSearch.
OpenSearch
- It is normally sufficient to have one cluster with at least two nodes, but three is recommended. The exception is if you have very large amounts of data to index and query, such as Promotions in the tens of thousands or content fragments in the hundreds of thousands.
- Combined with the recommendation to setup a backup mechanism, it may be required to set up a bigger cluster.
For complete details, refer to the following OpenSearch documentation:
- Scaling and deploying OpenSearch on a production environment:
- Backing up your indexes: https://www.elastic.co/guide/en/elasticsearch/guide/2.x/backing-up-your-cluster.html
Experience Optimization Management Service
One Experience Optimization Management Service per logical environment is usually sufficient. It is also possible to load balance two Management Services in the same logical environment.
Experience Optimization Query
Due to the high degree of variation in the data volumes and numbers of concurrent requests, SDL does not have hard metrics available on the number of Query Services that are needed on a production environment. At a minimum, we recommend having at least have two Query Services, load balanced to ensure high availability. For further scaling, you should conduct soak test to determine the expected load before going live with your scaled production query environment.