Probably in March began to surface several Internet companies, the main direction is the back-end java and large data development, recently put together learning materials, almost autumn recruited, hair a little late, but still want to share with you. US group, drops, mushrooms Street's face was.
Yoshi团
Side
chat you are most familiar items and technology.
Why use project spark.
how spark division stage, width dependence, chat source
spark submit a job execution process (stand-alone mode)
spark Driver node, worker and master node failures encountered how to solve.
spark checkpoint principle
chat JVM memory division and GC algorithm.
Why do I need to use the JVM generational collection algorithm, what are the benefits.
Shredded Code:
Enter a list and a value x, the value of the left requires replacement of all the nodes is less than x to x, and x is greater than all nodes to the right of substitution x, additionally required to maintain the original context. For example:
Enter: 1-> 4-> 3-> 2- > 5-> 2 and x = 3,
output: 1-> 2-> 2-> 3-> 4-> 5.
Wherein: 1/2 / 2 less than 3,4 / 5 greater than 3, and also before and after the order entered 1/2 / 2,4 / 5
side is just beginning distributed storage team, the main technology stack is hadoop, hbase and hive, then I feel distributed interest calculation, the interviewer helped me reorientation to real-time computing group, followed by two faces.
Two faces
self-introduction
chat program, which projects have difficulty to solve the problem?
How do spark streaming cluster disaster recovery process.
spark checkpoint principle.
What methods of mass data deduplication understanding.
flink and spark the difference?
flink understand what its basic architectural principle
of how to design a multi-level cache system, need to consider what issues?
redis and do zookeeper distributed lock distinction, there are other ways to do distributed lock on it?
An algorithm: Given a stack, the stack is calculated using the minimum element time complexity of O (1) is.
Which machine learning algorithms to understand? I was doing the project?
Teamwork between what issues need attention?
How do you do the same back to make a system, you'll learn how to experience from previous projects.
Three sides
introduce themselves
digging project.
How to understand big data, it solves the problem.
In a large array of m * n, each row ordered, disordered each column, find how its topk.
How to communicate between processes.
The concept operating system page, the size of each page is how much, and why are so many.
Whether a machine can establish a connection is unlimited, what factors affect it?
TCP time_wait off state when the connection?
5-tuple to determine a TCP connection.
What to ask my thing?
How to understand the real-time computing.
A few days later, asked the next, that is not the HC, I turn to distributed storage, but it needs to re-interview, no longer continue to face again.
Drops (through the interview)
一面
自我介绍
介绍项目。
说说JVM gc算法，垃圾回收器。
聊下java并发包的一些常用工具类
聊spark源码。
有什么想问的。
二面
redis的工作模型。
redis如何处理过期数据。
redis中hash类型求数据大小快还是将所有数据读出来快。
数据库与缓存读写一致性问题。
如果以上问题回答的不太完美，你会怎么办。
spark的通信模型，集群发生故障是怎么办。
spark的执行流程。
java中有哪些锁。
synchronized为什么是重量级锁。
如果工作中提出不合理的需求你会怎么办。
怎么看待加班。
实习时间。
过了几天面试官说面试通过，过几天老板会和我商量入职时间，然后过了几天。。。。HC被锁了。。
蘑菇街（offer）
Side
project description and difficult project and highlights
spark sql of the UDF, to achieve UDAF function.
Hive how to achieve UDF.
hashmap source in which the member variable, when it is appropriate to increase the load factor, when it is appropriate to reduce.
java a float type, type conversion.
Matching string.
Two faces
self-introduction
java in which a lock
java CAS, read the source code is not native method
What is a deadlock, the code appears how to solve the deadlock.
Seeking a binary tree topk large elements.
java volatile keyword.
spark execution flow.
How to solve the data skew.
Practice time, the tendency of the post.
Three sides
you have what are the advantages
impact on your college largest people
what strengths and weaknesses you have.
Interview on the front of the feel-how.
Expected salary
practice time.
Finally got the offer.
Summary: The dish is original sin! ! ! Mr. Lu said: No matter what, if continued to collect material, the plot of the decade, the total can become a scholar! Daily accumulation and precipitation is very important!
I remember the article points like a praise, thanks for the support!