Interview Review: Core Concepts of MySQL, OS, Networking, Redis, and Distributed Systems

Problem Records

MySQL - MVCC

InnoDB implements repeatable‑read isolation via MVCC, recording multiple versions of rows; transaction IDs, undo logs and versioning allow reading consistent snapshots.

MySQL - Atomicity

Atomicity is guaranteed by the undo log: before a row is modified, its original state is stored, enabling rollback of uncommitted transactions.

MySQL - Persistence

Persistence relies on the redo log and doublewrite buffer (WAL). Changes are first written to the redo log, then flushed to disk; the doublewrite buffer copies dirty pages before the final fsync.

Operating System - Deadlock

Deadlock occurs when two or more threads hold resources the other needs, forming a circular wait; it requires the four conditions of mutual exclusion, hold‑and‑wait, no preemption, and circular wait. Breaking any condition (e.g., ordered resource allocation) prevents deadlock.

Operating System - Page Cache

The page cache stores frequently accessed file blocks in memory, reducing kernel‑space copies. Reads and writes operate on the cached pages; modified pages are later flushed back to the file.

Computer Network - TCP Reliability and Ordering

TCP ensures reliable, ordered delivery using sequence numbers, ACKs, fast retransmit, and cumulative acknowledgments to detect and recover lost packets.

Computer Network - Flow Control

Flow control uses a sliding window. The receiver advertises its available buffer size; the sender limits transmission accordingly. The example shows a window of 200 bytes and how it shrinks and expands as data is sent and ACKed.

Redis - Persistence

Redis offers two persistence methods: AOF (append‑only file) records every write operation for exact reconstruction, while RDB creates point‑in‑time snapshots. AOF has lower runtime overhead but slower restarts; RDB restarts quickly but may lose recent data.

Redis - Clustering

Redis uses a master‑slave replication model for high availability and a hash‑slot sharding scheme for horizontal scaling; each key’s hash determines its slot and the node that stores it.

Distributed - Transactions

Distributed transactions are typically handled with two‑phase commit (prepare → commit) or three‑phase commit (CanCommit → PreCommit → Commit) to improve fault tolerance.

Distributed - Paxos vs Raft

Paxos allows followers to reject a leader’s proposal, requiring three phases; Raft simplifies consensus by electing a stable leader that all followers follow.

Distributed - Majority Consensus

Requiring a majority of nodes to agree prevents Byzantine failures and ensures that a single faulty node cannot compromise consistency.

Scenario Questions

Design a thread pool

Design an LRU cache

Interview Summary

The interviewer emphasized breadth over depth, covering many fundamentals without deep probing; the candidate noted the need to strengthen basic knowledge and make scenario questions more practical.