bulk insert b tree

[mehrdad3301]

When flusing a memtable, insertions are done one by one. A lot of splits and merges happen when the tree is built, and bulk inserts has much less IO cost. Do you think that implementing bulk inserts help in this case ? Does the memtable get large enough for asymptotic compexity improvement to take effect ?

[guycipher]

Batch writing would work so long as the sstable data is sorted.

[mehrdad3301]

Do you think it can lead to improvements ? Can I give it a shot ?

[guycipher]

I do! Absolutely, do report back your findings :)