Hi, I’m Johngo!
One month’s interview experience, 5 offers from big factories, and the final entry fast hand!
one
Into the job quickly
My brother, who accompanied me for several years in university, left his job again and went to another Internet company with large volume and large traffic – Kuaishou!
I’m really happy for him! We met in college, and we got jobs together when we graduated. All the way, it is not easy!
It has been nearly four years since I graduated, but I just stayed in one company and never had the experience brought by different environments. Probably more, the environment I’m in now is so familiar that I never think of leaving. Sometimes the heart will truly resist a new but unfamiliar environment.
But the reality is that job-hopping can lead to higher pay, more opportunities, and more challenges that can make your life experience more colorful. Of course, you may also lose something unexpected!
These unexpected things, after all, are different scenery on different roads. Encounter flowers, we feel the fragrance of flowers and abundant; Encountered spring rain, we trample mud against the wind, to meet the spring rain to bring all things recovery.
My friend, who is engaged in the same big data position as me, is the same direction we both took after graduation. This time, I probably asked for a frequent topic in the recent interview for his resignation.
Students who are ready or already on the job and want to be engaged in big data positions can refer to it. There are many topics that every company will involve. Therefore, instead of sorting out the output separately, most of the high frequency questions are listed.
two
High frequency problems
Divided into four modules
Flink related
1. How does Flink Exactly once? What’s the difference between Flink At Least Once and Exactly Once
2. How is Flink backpressure done?
3. How does Flink store state? What are the fault tolerance mechanisms?
4. What problems have YOU encountered when using Flink? What optimizations have been made?
5. The difference between Flink and Spark Streaming? What advantages does Flink have over Spark Streaming when handling real-time data? Why is that?
6. How did Flink Watermark work?
7. Types of Flink join, how interval Join two streams together? How to deal with the event of being late?
Hive related
1. How to solve data skew?
2. Will the window function in SQL be used? Name a few.
3. What are the common SQL optimization methods?
Kafka related
1. Why is Kafka fast in accessing data? What does page caching mean?
2. What steps does a Kafka producer go through before producing data?
3. What is Kafka ISR?
4. What does Kafka Controller do and how is it elected?
HBase
1. What is the process of reading and writing data?
2. Is the Hbase read or write efficiency high? Why is that?
3. What is the Hbase coprocessor working principle?
The above!
three
Big data r&d or data R&D?
Due to the fact that I was interviewed for flink-related positions, I might not have been asked about Spark. However, it is easy to notice that in some big factories, such as Bytecode, Kuaishou and Meituan, many of them are asked about the direction of data development, such as big data storage, SQL calculation and optimization.
After telling me about this phenomenon at that time, my first feeling was that dachang’s big data platform was probably mature, and various components were already running smoothly and smoothly. So the builders of big data platforms are stable and don’t need so many people to maintain them.
The rest is of course the upper data analysis and data application, to put it plainly, is the data support of flow realization, and these are in need of a large number of data research and development personnel to support. So, a large proportion of jobs now will be data development engineers. Will support algorithmic students, data analysts, etc.
four
LeetCode part
The last point, perhaps, is the algorithmic level of the interview you are concerned with (the underlying algorithm is not a machine learning algorithm).
This point is not summarized above, because the questions are rather general, and probably each interviewer has their own topic they want to ask, and this topic is often used in the work.
Or, in my soft and hard work, or to think of a roughly important sort:
A. string
B. Arrays and lists
C. Dynamic planning
d.DFS / BFS
E. trees and figure
F. other
But in my opinion, “tree and graph” is basically simpler than the others. Some of the ideas in “tree” can be used as the basic ideas of brush problems, which are very helpful for other algorithm modules such as dynamic programming.
In addition, it is mentioned that the algorithm brush problem group a period of time ago, the “tree” module of the first stage is about to enter the end, and there are friends together, can also join in. Private letter I ok!
In the actual interview, almost every company will be more or less mentioned in every round of interview, the only thing that is not mentioned is HR (manual dog head), LeetCode is still to be practiced as the main topic.
Some easy and some difficult, but byte and fast LeetCode is ok, the most difficult is medium.
Another saying is that people who work in these big factories have little time to work on these algorithm problems. So, they would prepare one or two algorithmic questions for everyone, year after year. Ha ha everybody understands, can explore the mouth wind in advance (…)
five
This time, I want to say, on the one hand, what is your opinion about frequent job hopping, and on the other hand, the direction of big data interview questions.
At the same time, I wish my classmates a bright future.
I also hope to meet all my classmates and colleagues who have helped each other and all of you who are reading this article.
Finally, we can not decide the direction of life and the final outcome, we can only go through setbacks and tribulations, keep moving forward. With firm goals and perseverance, the future must belong to all of us.
And that’s all for today!
Learn and know not enough, think and get foresight, action and can reach far.
See you next time!