ColinLeeo opened a new issue, #1074:
URL: https://github.com/apache/tsfile/issues/1074

   ## Problem
   
   `TsFileIOReader::get_timeseries_indexes` handles each requested measurement
   individually: binary search in the device index node, a separate
   `read_file_->read` (PREAD backend, so one syscall each), deserialization 
into a
   temporary arena, then a clone into the caller's arena. For a 4500-column 
table
   that is ~4500 preads plus ~4500 deserialize+clone.
   
   Profiling a wide-table read showed `do_load_timeseries_index` at ~7.6%.
   
   ## Proposed changes
   
   Collect every measurement leaf once (`get_all_leaf`), read the whole 
contiguous
   timeseries-index range with a single pread, and deserialize each index 
directly
   into the caller's arena; build a name -> index map and return the requested
   ones.
   
   Implemented in #1072; index loading dropped from ~7.8% to ~0.2% of the query.
   
   Related: #1072


-- 
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]

Reply via email to