Compare a B+ Tree storage engine (PostgreSQL / InnoDB) against an LSM-Tree storage engine (RocksDB / Cassandra) across Write Amplification, Read.

A strong answer

A strong candidate frames Compare a B+ Tree storage engine (PostgreSQL / InnoDB) against an LSM-Tree storage engine (RocksDB / Cassandra) across Write Amplification, Read. around the exact memory layout, compiler semantics, and hardware execution model covered in [LSM-Trees (MemTable & SSTables), Bloom Filters, and Leveled vs Size-Tiered Compaction]. Walk through the RUM Conjecture, contrast B+ Tree in-place 8 KB random page writes with LSM sequential SSTable flushes, prove why Bloom filters have 0% false negatives, and diagnose the LSM Tombstone range-scan anti-pattern.

Your IoT ingestion service writes 500,000 sensor telemetry records per second keyed by (device_id, timestamp) into a 20 TB database. On a B+ Tree engine, updating millions of scattered leaf pages across 20 TB triggers random 8 KB SSD read-modify-writes (Write Amplification Factor = 30–100x), wearing out SSDs and capping throughput at 20,000 writes/sec! How does switching to a RocksDB / Cassandra LSM-Tree absorb 500,000 writes/sec purely via sequential I/O—and how do Bloom Filters keep point reads from having to check 20 SSTable files on disk?

When engineers encounter this bottleneck in production, a surface-level API description rarely explains what went wrong. In real workloads, every line of code executes against concrete hardware constraints: fixed-width memory cells, CPU cache lines, kernel descriptor tables, network socket buffers, or accelerator memory buses. Without a clear mental model of Log-Structured Merge-Trees (LSM-Trees), MemTables, SSTables, Bloom Filters, and Compaction Amplification, debugging degrades into guesswork—tuning timeouts, adding retries, or scaling instances blindly while the underlying architectural flaw remains untouched.

In a production interview, always connect the specification rule to its concrete runtime consequence: state what lives on the stack versus the heap or static storage segment, how the compiler or operating system validates alignments, lifetimes, and error codes, and what happens under adversarial or boundary inputs (such as zero-length buffers, integer overflow, concurrent mutation, or partial I/O). Backing your explanation with a small trace of register/memory state transitions immediately separates senior systems engineers from candidates who only memorized syntax.

What the interviewer is listening for

• Mastery of core invariant: Trace the LSM-Tree write and read paths across the WAL, SkipList MemTable, and immutable SSTable levels.
• Mastery of core invariant: Prove why SSTable Bloom Filters (10 bits/key) have 0% false negatives and eliminate >99% of cold disk reads on point lookups.
• Mastery of core invariant: Balance Write, Read, and Space Amplification across Size-Tiered (STCS), Leveled (LCS), and Time-Window (TWCS) compaction.
• Hardware and ABI awareness: Explaining how sizes, alignments, calling conventions, or kernel boundaries behave on real 64-bit targets rather than relying on implementation-defined assumptions.
• Defensive engineering: Naming the exact compiler flags (-Wall -Wextra -Werror), static analyzers, or runtime sanitizers (-fsanitize=address,undefined,thread) that verify correctness in CI.

Common mistakes

• Explaining only surface syntax without tracing the underlying memory addresses, ownership lifetime, or kernel/user-space state transitions.
• Assuming undefined behavior (such as signed overflow, uninitialized reads, strict-aliasing violations, or dangling pointers) will "just work" because it passed an unoptimized -O0 debug build.
• Ignoring return codes, partial reads/writes, alignment padding, or cleanup paths when an allocation or syscall fails midway through a function.