Phone side

Introduce yourself, why you are studying for graduate school, etc
JVM memory structure, garbage collection.
Java concurrency
- What are the
  reference answers for Java locks : fair lock/unfair lock; whether it is reentrant; exclusive lock/shared lock, read-write lock ReentrantReadWriteLock can be regarded as a combination of exclusive lock and shared lock; pessimistic lock/optimistic lock (CAS)
- The difference between synchronized and lock
Architecture and optimization
- General idea
- Types of load balancing: layer four, based on ip+port; layer seven, based on url; DNS (refer to the summary description of linux load balancing (layer four load/layer seven load) ). Load balancing algorithm, consistent hash
- The difference between mysql master-slave and master-slave. The main storage engine of mysql and its differences. Affairs.
- Distributed transaction: please refer to 4.4.5 for detailed explanation of distributed consensus protocol and section 6 distributed transaction answer
- Distributed locks
  I mentioned that redis and zookeeper can be used to implement distributed locks, but he also asked how to implement it with a database. You can refer to distributed locks based on database.
- Database and cache consistency

Two sides of the phone

Self-introduction etc.
Core parameters of thread pool
Lock upgrade, lightweight lock. How does CAS solve the ABA problem. volatile.
JVM heap area structure. How to tune. The advantages of G1. How to troubleshoot online OOM issues. How to troubleshoot if CPU load is too high? What is your first impression? Answer: code problems such as infinite loop. Q: Is GC possible? ¹
Some key knowledge points of NIO ² , ³ , ⁴
:
BIO data read and write must be blocked in a thread waiting for its completion.
IO multiplexing: one thread can handle multiple network connections.
Selector: no longer polling channels, but polling events. Based on event-driven mechanism. epoll.
Direct memory can be manipulated in NIO, which is also a reason for its better performance than BIO.
^{In 5} , an IO operation is divided into two steps: initiating an IO request and the actual IO operation to distinguish whether it is blocked, and synchronous/asynchronous. I think it seems to help understanding.
Netty ⁶
Sub-library and sub-table
Distributed transactions. TCC
Primary key index
Are you familiar with big data products? For example, stream computing, like Spark, Flink.
zookeeper electoral process ⁷ . Implement distributed lock with zk
Kafka throughput is relatively high. Message sequence.
Spike