Interview Prep · Phase 4
The Java Memory Model and everything built on it: happens-before, why volatile is not a lock, how to size a thread pool from measurements rather than folklore, composing CompletableFuture chains, and diagnosing a deadlock from a thread dump.
The six thread states and what moves between them, why calling run() directly is a no-op, what a platform thread actually costs in memory and switching, and the interrupt protocol done properly.
Read tutorialWhat the Java Memory Model guarantees, why reordering and visibility are separate problems, what synchronized actually does beyond mutual exclusion, and when ReentrantLock earns its extra complexity.
Read tutorialWhat volatile guarantees and what it does not, why count++ is broken even when volatile, how compare-and-swap works, when LongAdder beats AtomicLong, and the double-checked locking idiom.
Read tutorialHow ThreadPoolExecutor decides whether to queue or grow, why newFixedThreadPool can exhaust the heap, sizing pools from measurements with Little law, and shutting down without losing work.
Read tutorialComposing async work without blocking: thenApply versus thenCompose, which thread runs each stage, combining with allOf and anyOf, timeouts, and how exceptions propagate through a chain.
Read tutorialThe four conditions every deadlock needs and how breaking one prevents it, reading a deadlock out of a thread dump, lock ordering and tryLock, and the pool-starvation deadlock with no locks at all.
Read tutorialThe coordination primitives in java.util.concurrent, when a latch beats a barrier, using a semaphore as a bulkhead, and the AbstractQueuedSynchronizer that all of them are built on.
Read tutorialHow a virtual thread mounts and unmounts from a carrier, why pinning on synchronized still matters, why pooling virtual threads is wrong, and what StructuredTaskScope adds over raw futures.
Read tutorial