I wonder how much RAM did you have on the system where this test was performed? Check https://bugs.mysql.com/bug.php?id=69391 - table_definition_cache is one of the variables that defines the amount of memory allocated for the performance_schema, and maybe it was just swapping with high value?
Also, is it really important to have 700K tables to reproduce the notable difference in startup time with P_S set to ON vs OFF with table_definition_cache=1000 (it seems so from perf output, but just to double check)?
I wonder how much RAM did you have on the system where this test was performed? Check https:/ /bugs.mysql. com/bug. php?id= 69391 - table_definitio n_cache is one of the variables that defines the amount of memory allocated for the performance_schema, and maybe it was just swapping with high value?
Also, is it really important to have 700K tables to reproduce the notable difference in startup time with P_S set to ON vs OFF with table_definitio n_cache= 1000 (it seems so from perf output, but just to double check)?