Roadmaps / Fundamentals / Interview topics
Reference
Interview topics
Use these for backend, systems and infrastructure roles. Each follow-up is the question a good interviewer asks next. The Staff line is what lifts the answer.
Phase 1
| Question | Follow-up | A Staff answer adds |
|---|---|---|
| What happens when you allocate memory? | When does it become slow? | The allocator, the collector or the ownership model and a measured cost |
| How do you find a memory leak? | What about in a garbage-collected language? | A heap profile, a growth rate and the retaining path |
| Why are benchmarks misleading? | How do you make one fair? | Warm-up, repeats, spread and a statement of what it does not show |
| How do languages differ on errors and nulls? | What bugs follow? | Examples from two languages and a team rule |
System design prompt (20 minutes): a service that keeps 100,000 sessions in memory. Outline: requirements and numbers (3 min), data layout and memory per session (5), eviction (4), failure and restart (4), observability (4).
Phase 2
| Question | Follow-up | A Staff answer adds |
|---|---|---|
What does fsync guarantee? | What can still lose data? | Directory syncs, device caches and a crash test that proves the claim |
Threads, epoll or io_uring? | How do you decide? | The connection profile, the memory budget and a measurement |
| How do you find why a process is slow? | What if CPU is low? | Off-CPU analysis, system call traces and I/O latency histograms |
| What happens in a container with a memory limit? | What kills the process? | The cgroup limit, the out-of-memory killer and how to see it |
System design prompt (20 minutes): a write-ahead log for a small database. Outline: record format (4 min), flush policy (5), recovery (5), testing (4), cost (2).
Phase 3
| Question | Follow-up | A Staff answer adds |
|---|---|---|
| What is a data race? | How do you find one? | The memory model, a detector and what the detector cannot see |
| How do you avoid deadlock? | How do you debug one in production? | A lock order, timeouts and a thread dump taken before restart |
| Lock, atomic or channel? | What does each cost? | The invariant it protects and a measured cost |
| What is false sharing? | How do you detect it? | Cache lines, a counter experiment and when it is not worth fixing |
System design prompt (20 minutes): a work queue shared by 16 threads. Outline: requirements (3 min), queue design and bounds (5), correctness tools (4), failure and shutdown (4), measurement (4).
Phase 4
| Question | Follow-up | A Staff answer adds |
|---|---|---|
| What happens when you type a URL? | Where can it fail? | DNS, TCP, TLS and HTTP stages, each with a tool to observe it |
| What is a half-open connection? | How do you handle slow clients? | Timeouts, keepalive, bounded buffers and a test |
| How do you read a packet capture? | What would a reset tell you? | The sequence of packets, who sent the reset and the likely causes |
| How do you size a server? | How do you check the estimate? | A load test with percentiles, headroom and a stated failure point |
System design prompt (20 minutes): a gateway that holds 100,000 idle connections. Outline: numbers (3 min), memory per connection (5), timeouts and limits (4), failure modes (4), observability (4).