taberylyang commented on pull request #8393: URL: https://github.com/apache/incubator-doris/pull/8393#issuecomment-1062603382
> I have a question If there is a broadcast join, the hash table of each node is actually constructed concurrently at the same time, so the construction time will not affect the query. On the contrary, it will take longer to build a single node first and then share the hash table. So why can q4.1 in ssb improve 3s? > > 我有一个疑问，如果是 broadcast join 那么各个节点的hash table 其实是同时并发构建的，所以构建的时间并不会影响查询。反而是，如果先构建单个节点再分享hash table 的时间会更久。那么 ssb 中的 q4.1 为何能提升3s 呢？ 我这边是这么测试的，单be下使用空的含有32个tablet的lineorder表作为内表，q4.1的其他表不变。然后执行q4.1，由于lineorder是空表，所以可以认为所有时间都消耗在了语句的open阶段，测试结果是需要3s左右。当我降低空的lineorder的tablet数量，会发现语句执行时间明显缩短，最多只需要0.5秒以内。通过perf也能发现，时间都消耗在了构建哈希表。 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
