Which statement best describes a Log-Structured Merge Tree (LSM Tree)?

Which statement best describes a Log-Structured Merge Tree (LSM Tree)?

• A. A spatial indexing approach that converts 2D coordinates into a 1D string where nearby locations share similar string prefixes.
• B. A write-optimized storage format that batches writes in memory and flushes them to disk sequentially, trading reads for faster writes.
• C. A persistent hashmap that maps indexed values directly to row locations for O(1) lookups.
• D. A versioned log system used for auditing.

Answer: (B)

LSM trees optimize writes by buffering them in memory and flushing them to disk in large, sequential chunks. Writes go first to an in-memory structure (a memtable), which allows very fast, random-write performance. When the memtable fills, its contents are written out to disk as sorted runs, leading to efficient, sequential disk I/O. Over time, these on-disk runs are merged through compaction to keep the dataset organized and to reclaim space. Reads become a bit more involved because they may need to check multiple levels or structures to locate a key, which is the trade-off for achieving high write throughput. This matches the idea of a write-optimized storage format that batches writes in memory and flushes them to disk sequentially, trading reads for faster writes. The other descriptions describe different data structures or logging concepts that aren’t how LSM trees operate.

LSM trees optimize writes by buffering them in memory and flushing them to disk in large, sequential chunks. Writes go first to an in-memory structure (a memtable), which allows very fast, random-write performance. When the memtable fills, its contents are written out to disk as sorted runs, leading to efficient, sequential disk I/O. Over time, these on-disk runs are merged through compaction to keep the dataset organized and to reclaim space. Reads become a bit more involved because they may need to check multiple levels or structures to locate a key, which is the trade-off for achieving high write throughput. This matches the idea of a write-optimized storage format that batches writes in memory and flushes them to disk sequentially, trading reads for faster writes. The other descriptions describe different data structures or logging concepts that aren’t how LSM trees operate.