讨论最近面试一道题：design a action events collection workflow

最近刚面了汉宾罗的system design店面。楼主觉得答的不错，但是还是挂了。想请教一下地里的小伙伴能否提供一些新鲜思路。

Users from the XX client (IOS, Android, Web) can generate arbitrary actions (click some button, browse some page, etc.). Those events with timestamp can be useful for XX to learn the user behaviors. XX either can do the analysis or build some dashboard on top of these events data.  Please design a workflow system to collect these events.

我自己的想法是用到了load balancer, a set of receiver servers, message queue, a pool of workers. 然后把data dump到sharded and distributed files with index (no need database). 上层有相应的query engine to query it.

放到object store上？需要实时的么？

我觉得更重要的是数据结构：ID，ts，action ID都要能方便的group。可能还需要支持transaction。所以你需要设计一个分布式数据库来存这些数据。

我好像在某卵上也被问了？？？

补充内容 (2020-12-17 06:38):
Um... 仔细一看 不是一样的