The basic implementation used today to record metrics (A/B/n and performance) is an in memory key-value store implemented by badgerdb. This implementation should be replaced with a persistent data store that scales horizontally. There is not need that it be a key-value store. Another form of database is possible. The requirements are multiple writers with high write throughput. Multiple readers with low read throughput (once per 30 seconds by grafana per test).
The basic implementation used today to record metrics (A/B/n and performance) is an in memory key-value store implemented by badgerdb. This implementation should be replaced with a persistent data store that scales horizontally. There is not need that it be a key-value store. Another form of database is possible. The requirements are multiple writers with high write throughput. Multiple readers with low read throughput (once per 30 seconds by grafana per test).
Related to https://github.com/iter8-tools/iter8/issues/1480