LebronAl edited a comment on pull request #4570: URL: https://github.com/apache/iotdb/pull/4570#issuecomment-994912342
> 1. Adding schemas to GetAllPathsResult may double its size, which may cause performance issues. So my suggestion is to provide 2 RPCs, one for basic information like pathname, data type, alias, and the other one for extended information like encoding and compression type. I don't think adding the Encoding will double the size of GetAllPathsResult. In my opinion, most of the space in this structure is taken up by paths and aliasList of type List< String>, and List< byte> should not take up much space. If you really want to save it, it's possible to maintain an additional RPC interface, but you can simply add parameters to the current RPC interface to indicate whether additional fields are needed. In the current cluster architecture, our CMManager will transfer a large number of paths in RPC, which is relatively performance adverse. Our recently designed new cluster architecture will try to avoid this process, welcome to join the discussion. > 2. For SingleSeriesQueryRequest , since the coordinator may not have the metadata, it is reasonable to let the data group member set the schema. Same for MultSeriesQueryRequest. If the coordinator node doesn't need to retrieve metadata to build queryPlan, then I don't think there's anything wrong with your statement. However, in the current architecture, the coordinator node already needs to fetch all the metadata when parsing SQL into a Queryplan. In my opinion, rather than having the data node send the RPC again to fetch the metadata, it would be better to have the RPC transmit a few more bytes (most of which are occupied by the string variable). What do you think? -- 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]
