请问HN:处理用户可见分析的最佳方法是什么?
我正在寻找关于用户可视化分析功能的技术策略建议(比如实时可定制的仪表板)。我目前的情况是,想为最终用户设计一些分析仪表板,以显示多年的汇总/历史数据。我想运行的查询在实时执行时耗时太长,因此我不得不定期预计算我关心的数据。这种方法虽然有效,但并不是实时的,并且在引入新功能/视图时增加了一些阻力(每个新功能都需要填补多年的数据)。我想知道是否有其他人有更好的策略来开发这样的分析软件,使其能够实时响应,并且便于快速迭代?比如使用像 ClickHouse 这样的 OLAP 数据库?反向 ETL?更好的 OLTP 索引?
查看原文
I’m looking for advice on technical strategies for user-facing analytics features (think real-time customizable dashboard). I’ve ended up in a situation where I want to design some analytics dashboards for end users that display aggregates/historical data for a number of years. The queries I want to run take way too long to run on-the-fly so I’ve resorted to pre-computing the data I care about on a schedule. This approach works fine but it’s not real time and it adds some friction to introducing new features/views (every new feature requires backfilling years worth of data). I’m curious if others have better strategies for developing analytics software like this that’s real-time, responsive, and quick to iterate on? OLAP DBs like clickhouse? Reverse ETL? Better OLTP indexing?