wiki:Database Optimization

Version 1 (modified by crschmidt@…, 14 years ago) (diff)


09:24:58 * Dom recommends inspecting 09:27:10 <@crschmidt> Dom: the problem is that the metadata table doesn't have a useful index -- 09:27:52 <@crschmidt> the most common use cases are "Select all metadata for a node, given an id and a version", and "Select all nodes, given a metadata_type and


09:28:24 <@Dom> crschmidt: isn't that what, say, is doing? 09:28:52 <@crschmidt> No, that creates a single index -- those can't be split up 09:29:07 <@Dom> ah, right. 09:29:30 <@PerfDave?> Raw SQL rather than DBIx::Class? ;) 09:29:51 <@crschmidt> Creating an index on node_id,version would speed up the first case (I Think) and creating an index on metadata_name, metadata_value would speed

up the second (I think)

many "I Think, but am not sure" disclaimers throughout.