Du Xiaoman Java Backend Salary & Interview Secrets Revealed

Salary Overview

Du Xiaoman's 2023 campus recruitment for backend development offers total compensation comparable to major internet firms, ranging from 36 w to 42 w per year. The packages include a base salary (e.g., 24k, 26k, 27k, 28k) multiplied by 15 months, with additional signing bonuses of up to 7 w, paid in two installments.

Base 24k × 15 months → 36 w total

Base 26k × 15 months + 7 w signing → 39 w total

Base 27k × 15 months + 7 w signing → 40.5 w total

Base 28k × 15 months → 42 w total

The signing bonus functions as a retention incentive; if an employee leaves before receiving the second installment, the remaining amount is forfeited.

Internship Value

An intern who secured a Du Xiaoman position later received a 28k offer from a top-tier company, demonstrating the career boost that a reputable internship can provide.

Interview Preparation

Candidates are advised to start early, polish algorithm skills, and prepare project portfolios before the summer recruitment wave begins.

1. Thread Pool Mechanics

Thread pools limit the number of concurrent threads, improve resource management, and reduce the overhead of frequent thread creation and destruction.

Core thread count is configurable.

Idle threads are retained for quick task execution.

Facilitates monitoring and tuning.

The execution flow is:

Check pool state; reject if not RUNNING.

If workerCount < corePoolSize, create a new thread.

If workerCount ≥ corePoolSize and the queue is not full, enqueue the task.

If the queue is full and workerCount < maximumPoolSize, create a new thread.

If the pool is at maximum size and the queue is full, apply the rejection policy (default: throw exception).

Key parameters:

corePoolSize : Minimum number of threads kept alive.

maximumPoolSize : Upper limit of threads.

keepAliveTime : Time after which excess idle threads are terminated.

unit : Time unit for keepAliveTime.

workQueue : Queue that holds pending tasks.

threadFactory : Factory for creating threads with custom names or priorities.

handler : Rejection strategy when the pool cannot accept new tasks.

2. synchronized vs. ReentrantLock

Usage : synchronized can annotate methods, static methods, or blocks; ReentrantLock works only on code blocks.

Lock acquisition : synchronized acquires and releases automatically; ReentrantLock requires explicit lock() and unlock().

Fairness : synchronized is non‑fair; ReentrantLock can be configured as fair or non‑fair.

Interruptibility : ReentrantLock supports interruption; synchronized does not.

Implementation : synchronized relies on JVM monitors; ReentrantLock is built on AQS.

3. Other Lock Implementations

ReentrantReadWriteLock : Provides separate read (shared) and write (exclusive) locks.

StampedLock : Offers write lock, pessimistic read lock, and optimistic read lock for high read‑write ratio scenarios.

Condition : Used with Lock to enable advanced thread coordination.

Semaphore, CountDownLatch : Synchronizers based on AQS, not direct Lock implementations but frequently used.

4. AbstractQueuedSynchronizer (AQS) Basics

AQS provides a framework for building synchronizers using a volatile state variable and a CLH queue of waiting nodes. Threads attempt a CAS on state to acquire; on failure they enqueue and block. AQS supports exclusive and shared modes, handling the low‑level blocking and wake‑up logic.

5. Compare‑And‑Swap (CAS) Principle

CAS atomically compares a memory location V with an expected value A; if equal, it writes a new value B. It is the foundation of Java's Unsafe operations and atomic classes like AtomicInteger.

6. CAS Pitfalls & Solutions

ABA problem : Resolved with versioned references such as AtomicStampedReference.

Excessive spinning : Limit spin retries or fall back to traditional locks.

Limited to single‑variable atomicity : Combine multiple fields into a single object and use AtomicReference, or protect with locks.

7. SQL Optimization Techniques

Analyze queries with EXPLAIN to identify full scans.

Create appropriate indexes (single‑column, composite, respecting left‑most rule).

Avoid index‑breaking patterns (leading wildcards, functions on indexed columns).

Prefer covering indexes and limit SELECT * usage.

For deep pagination, use WHERE id > last_id LIMIT n instead of LIMIT offset, n.

Implement read‑write separation with master‑slave replication.

Introduce caching layers (e.g., Redis) while handling cache consistency.

8. Sharding: Database & Table Partitioning

Sharding splits data across multiple databases or tables to handle large volumes and high concurrency.

Database sharding : Distributes load across separate DB instances.

Table sharding : Divides a large table into smaller tables with identical schema.

Vertical sharding : Separates infrequently used or large columns into a different table.

Horizontal sharding : Splits rows based on a key (e.g., user ID, time) into multiple tables.

9. Handling Aggregation After Horizontal Sharding

Select an appropriate sharding key to keep most queries single‑shard.

If cross‑shard aggregation is unavoidable, perform it in the application layer or use middleware like ShardingSphere.

For complex analytics, offload data to Elasticsearch or a data warehouse.

10. Transaction Isolation Levels

Read Uncommitted

Read Committed

Repeatable Read (default for MySQL InnoDB)

Serializable

11. MVCC Mechanism

Multi‑Version Concurrency Control allows concurrent reads without blocking writes by maintaining versioned rows. In InnoDB each row stores trx_id (creating transaction ID) and roll_pointer (pointer to the previous version). A Read View captures min_trx_id, max_trx_id, and the list of active transaction IDs ( m_ids) to determine visibility of each version.

12. Sample Algorithm Question

One of the interview’s algorithm tasks asks for the maximum number of points on a straight line.

Overall, the article combines salary transparency, internship impact, and a comprehensive technical guide to help candidates succeed in Du Xiaoman’s Java backend recruitment process.