5.9 KiB
5.9 KiB
1 - What are atomic variables?
Explain their purpose and how they differ from ordinary (non-atomic) variables.
Answer:
Atomic variables are variables on which reading and writing operations are atomic.
It means that when a Thread is making changes to a variable, no other Thread can interfere with it.
They Use hardware-level instructions to ensure that only one thread can modify the value at a time without locks.
But for ordinary variables, operations like balance += amount are not atomic for example if another thread interrupts between these steps, it causes Race Conditions and data corruption.
2 - Name at least four classes from the java.util.concurrent.atomic package that provide atomic operations for different data types.
For one of them, briefly describe a typical use case.
Answer:
- AtomicInteger
- AtomicLong
- AtomicBoolean
- AtomicReference
Example :
private AtomicLong balance = new AtomicLong(0);
public void deposit(long amount)
{
balance.addAndGet(amount);
}
Atomically adds amount to current balance.
It is fast & it does not need synchronized
.
But it is only good for simple operations on one variable.
3 - Compare locks with atomic variables.
In which scenarios is using a lock a better choice than an atomic variable, and vice versa?
Answer:
| Feature | Lock (ReentrantLock / synchronized) | Atomic Variables |
|---|---|---|
| Complexity of operations | Great for complex, multi-step operations | Great for simple operations on a variable |
| Scope of Locking | You can lock blocks of code | It only locks the variable itself |
| Conditional Logic | Supports conditional checks;You can check if (balance > 0) and then withdraw | Difficult; requires loops with compareAndSet which can be complex. |
| Resource Management | Requires manual unlocking in finally | Automatic |
| Deadlock | Dangerous | None |
When to Use Which?
Use Locks When:
- You need to perform compound operations involving multiple state variables atomically (e.g., in our banking project: balance -= amount AND target.balance += amount must happen together).
- You need to wait for a condition (using Condition.await()/signal()).
- The critical section contains complex logic that cannot be reduced to a single atomic instruction.
Use Atomic Variables When:
- You are modifying a single variable (e.g., a global counter, a status flag).
- Performance is critical and contention is expected to be low.
- You want to avoid the complexity of managing lock lifecycles (acquire/release).
4 - A program is completely free of race conditions but still performs poorly under high contention.
Explain how this situation can occur.
Discuss at least three concurrency-related factors that may limit scalability even when correctness is guaranteed.
Answer:
Even without race conditions, high contention causes bottlenecks due to:
- Thread Contention: Threads block waiting for locks, turning parallel execution into sequential processing. CPU time is wasted managing queues rather than computing.
- Context Switching Overhead: Frequent blocking/unblocking forces the OS to save/restore thread states. High switch rates consume CPU cycles needed for actual work.
- False Sharing: Unrelated variables in the same cache line cause unnecessary cache invalidations across cores, forcing slow main memory accesses despite logical independence.
5 - Many concurrent systems experience performance degradation as the number of threads increases.
Explain why adding more threads does not always improve performance.
Your answer should discuss concepts such as:
Context switching
Contention
coherence
Synchronization overhead
Answer:
Performance degrades with excessive threads due to:
- Synchronization Overhead: Lock acquisition/release costs exceed computation time for small tasks.
- Contention: Increased probability of lock conflicts leads to long wait times, shifting from parallel to serialized execution.
- Context Switching: Beyond core limits, CPUs spend more time switching contexts than executing instructions (“thread explosion”).
- Cache Coherence: Frequent writes to shared data trigger inter-core synchronization (MESI protocol), saturating the communication bus and increasing latency
6 - Deadlocks often only appear in production, not during testing.
Explain why this might happen from a thread-scheduling perspective.
Describe two strategies a developer can use to increase the likelihood of exposing deadlocks during testing.
Answer:
Why they hide in testing:
- Timing: Tests are fast and deterministic; production has I/O/network delays that alter thread interleaving.
- Concurrency: Tests use few threads; deadlocks often require high concurrency to trigger naturally.
Strategies to Expose Deadlocks:
- Stress Testing: Spawn hundreds of threads performing random operations to exponentially increase the chance of hitting cyclic dependency windows.
- Inject Delays: Add Thread.sleep() or use tools like JMH to disrupt natural flow, increasing the likelihood that a thread holds one lock while waiting for another.