At present, in addition to the raw data query without value filtering, the processing of vector time series is to convert it into components and then treat it as an ordinary time series, that is, if K components of vector time series are queried, K readers need to be built for K queries. Therefore, our next optimization is, group K components belonging to the same vector time series for query, that is, query only once to find out all component data. In aggregation, it involves how to use the statistical information of multiple components and how to map multiple component value columns in batchdata to each component for update.
I've designed it briefly in [1]. If you have interest in it, please have a look~ [1] https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=188746001 -- Best, Xiangwei Wei
