> bloom filters can guess right sstables to be read with high > probability < 0.1%. In reality even if you are using size based > compaction and have about 300 sstables, reading is fast unless > there is row fragmentation and you are reading entire row.
Right, that's it. Thanks!
