Hi, Recently I am checking the performance of creating tables.
I found that MigrationStage thread spent most of the time in " mergeColumnFamilies(oldColumnFamilies, newColumnFamilies);" method. Can someone explain the purpose of merging oldCFs and newCFs to find the diff between old and new when C* already knew which table should be created? Is it expected behaviour? If C* directly applied the new table CFMetate without merging-function, the performance of creating table is improved significantly. Mechina: OSX, Intel I5, 12GB Ram, SSD, One Cassandra Node. 2.1.14 creating 1500 table: 21m. 2.1.14 creating 1500 table without mergingCF : 9 m. When table number grows, the time difference gets larger. The attachment is the CPU time of MigrationStage in "mergeColumnFamilies( oldColumnFamilies, newColumnFamilies);". Thank you. Jasonstack [image: cpu_time.png]
