Concurrency Interview

What is the difference between concurrency and parallelism, and can you have one without the other?

Two threads each run `counter++` a million times on a shared integer. The final value is less than two million. Explain exactly why, and what your options are.

When do you use a mutex, and how do you decide what goes inside it?

When would you choose async/await over threads, and when would you choose threads?

Walk me through how a deadlock occurs and how you would prevent one in a system with several locks.

A service is slow. Someone suggests doubling the thread count. Will that make it faster?

What is the difference between a race condition and a data race? Can you have one without the other?

You replace a mutex-protected integer with an atomic. What have you gained, and what have you not?

Thread A writes a struct and then sets a flag. Thread B sees the flag and reads the struct. What can go wrong?

How many threads should a pool have?

A producer generates work faster than a consumer can handle it. What are your options, and what does each cost?

A client disconnects while their request is being processed. What should happen to the work in flight?

Someone proposes rewriting a contended queue as a lock-free structure for performance. How do you respond?

You parallelise a job across eight cores and get a 3x speedup. Why, and what do you do about it?

What is the GIL, and what does it mean for concurrency in Python?

Every endpoint on a Node service gets slower at the same time, including a health check that does nothing. What do you look for?

Two users edit the same record. Would you use optimistic or pessimistic concurrency control, and how do you decide?

Four threads each increment their own counter in an array. There is no lock and no shared variable, and the parallel version is slower than the sequential one. Why?