If your writes are larger than ~1/2 a page, LSMs are slower, regardless of random or sequential access pattern. http://lmdb.tech/bench/ondisk/ http://lmdb.tech

If your writes are larger than ~1/2 a page, LSMs are slower, regardless of random or sequential access pattern. http://lmdb.tech/bench/ondisk/ http://lmdb.tech/bench/ondisk/ Also, if your writes are mostly smaller than ~1/2 a page, you can reduce your B+tree pagesize and regain performance.