What is the role of a Bloom filter in LSM-tree storage?

What is the role of a Bloom filter in LSM-tree storage?

• A. A deterministic data structure that guarantees presence of keys.
• B. A probabilistic data structure attached to each SSTable that can definitively rule out whether a key is absent.
• C. A cache of recently accessed keys.
• D. A structure used to reorder keys on disk.

Answer: (B)

In an LSM-tree, the Bloom filter is a small, fast, probabilistic data structure attached to each on-disk SSTable. Its job is to quickly answer a simple question: could this SSTable contain the key I’m looking for? If the Bloom filter says no, you can conclusively skip reading that SSTable entirely, which saves a disk I/O. If it says maybe, you still have to check the SSTable to confirm. The key idea is that Bloom filters never give false negatives (if the key is actually in the SSTable, the filter will not rule it out), but they can give false positives (the filter may indicate the key could be present even when it isn’t), which means you might perform an unnecessary disk read in that case. Overall, Bloom filters dramatically reduce unnecessary disk access by pruning away SSTables that cannot contain the target key.

In an LSM-tree, the Bloom filter is a small, fast, probabilistic data structure attached to each on-disk SSTable. Its job is to quickly answer a simple question: could this SSTable contain the key I’m looking for? If the Bloom filter says no, you can conclusively skip reading that SSTable entirely, which saves a disk I/O. If it says maybe, you still have to check the SSTable to confirm. The key idea is that Bloom filters never give false negatives (if the key is actually in the SSTable, the filter will not rule it out), but they can give false positives (the filter may indicate the key could be present even when it isn’t), which means you might perform an unnecessary disk read in that case. Overall, Bloom filters dramatically reduce unnecessary disk access by pruning away SSTables that cannot contain the target key.