CareerCup

You have a huge file that contains 3 columns employee id, department id and salary. Give me the avg salary per department. I said I would use a hash table.
He asked what are the performance issues with this appraoch.
Then he said given that you have multiprocessor system available what can you do. I said we can open multiple filepointers for reading from different parts of the file and write each record into a hash table which is made thread safe. He said that would help only if the file is stored in different tracks.

Then he asked, for every record you are writing into the hashtable waiting for a lock on it and writing. This is very inefficient, tell me how you can make it more efficient. I said I will use a cache in each of the thread and write it out to the hashtable only when cache is full. He said that is a good solution.