Scalability, Availability & Reliability Interview Questions Hindi mein

Scalability, Availability & Reliability कोर्स के 25 सवाल, हर सवाल के सही जवाब और explanation के साथ। जवाब देखने से पहले ख़ुद सोचें।

Definitions, Targets and the Maths Behind Them

1. An API responds quickly to every request but sometimes returns stale, incorrect balances. Which property is weak?

  1. Reliability (correctness)
  2. Availability
  3. Scalability
  4. Throughput
जवाब देखें

जवाब: Reliability (correctness)

It is available and fast, but not doing the right thing consistently.

पाठ: Scalability, Availability and Reliability

2. With a 99.9% monthly availability SLO and 10 million requests, how many failed requests does the error budget allow?

  1. 100
  2. 1,000
  3. 100,000
  4. 10,000
जवाब देखें

जवाब: 10,000

0.1% of 10,000,000 is 10,000 requests.

पाठ: SLIs, SLOs, SLAs and Error Budgets

3. A service receives 500 requests per second and each takes 200 ms on average. How many requests are in flight on average?

  1. 2.5
  2. 700
  3. 100
  4. 2,500
जवाब देखें

जवाब: 100

Little's law: L = 500 × 0.2 = 100.

पाठ: Little's Law, Amdahl's Law and Queueing

Scaling Compute

4. Why must a service usually be stateless to scale horizontally?

  1. Stateless code runs faster on one CPU
  2. Load balancers reject stateful services
  3. So any instance can serve any request, and instances can be added or removed freely
  4. Stateless services do not need databases
जवाब देखें

जवाब: So any instance can serve any request, and instances can be added or removed freely

If state lives in one instance, requests depend on reaching that particular instance.

पाठ: Vertical vs Horizontal Scaling and Statelessness

5. Request costs vary widely: some take 5 ms, others 2 seconds. Which algorithm usually balances better than round robin?

  1. Round robin
  2. Least outstanding requests (or least connections)
  3. Alphabetical by hostname
  4. Always the first server
जवाब देखें

जवाब: Least outstanding requests (or least connections)

Load-aware algorithms avoid piling requests onto an instance busy with slow work.

पाठ: Load Balancing

6. Which metric is usually best for autoscaling background workers that read from a queue?

  1. Queue depth or age of the oldest message
  2. Disk size
  3. Number of deployments per day
  4. Memory of the load balancer
जवाब देखें

जवाब: Queue depth or age of the oldest message

Queue backlog directly reflects how far workers are behind demand.

पाठ: Autoscaling

Scaling Data

7. A user edits their name and the next page still shows the old name. What is the likely cause?

  1. The read was served by an asynchronous replica that had not caught up
  2. The leader rejected the write
  3. Synchronous replication
  4. The load balancer cached the name forever
जवाब देखें

जवाब: The read was served by an asynchronous replica that had not caught up

Replica lag causes stale reads right after a write; read-your-writes routing fixes it.

पाठ: Replication

8. What is the main advantage of consistent hashing over hash-modulo-N?

  1. It supports SQL joins across shards
  2. It eliminates hot keys entirely
  3. It keeps keys in sorted order
  4. Adding or removing a node moves only a small fraction of keys
जवाब देखें

जवाब: Adding or removing a node moves only a small fraction of keys

Only keys adjacent to the changed node on the ring move.

पाठ: Partitioning and Sharding

9. With N = 3 replicas, which setting guarantees reads overlap the latest write?

  1. W = 1, R = 1
  2. W = 1, R = 2
  3. W = 2, R = 2
  4. W = 0, R = 3
जवाब देखें

जवाब: W = 2, R = 2

R + W = 4 > 3, so the read set always includes a replica with the latest write.

पाठ: Consistency Trade-offs: CAP, PACELC and Quorums

Performance Under Load

10. Many requests miss the cache at the same moment when a popular key expires and all hit the database. What is this called?

  1. Cache warming
  2. Write-through caching
  3. Cache stampede (thundering herd)
  4. Replica lag
जवाब देखें

जवाब: Cache stampede (thundering herd)

Simultaneous misses on a hot key overload the backing store; coalescing and jitter help.

पाठ: Caching as a Scaling Tool

11. What does back-pressure achieve in an overloaded pipeline?

  1. It makes the queue infinitely large
  2. It signals producers to slow down so the backlog and memory stay bounded
  3. It deletes the oldest data silently
  4. It disables batching
जवाब देखें

जवाब: It signals producers to slow down so the backlog and memory stay bounded

Back-pressure propagates overload upstream instead of letting queues grow without limit.

पाठ: Queues, Load Levelling and Back-Pressure

12. Why should latency SLOs use percentiles such as p99 rather than averages?

  1. Averages hide the slow requests that many users experience, especially with fan-out
  2. Averages are harder to compute
  3. p99 is always lower than the average
  4. Percentiles ignore errors
जवाब देखें

जवाब: Averages hide the slow requests that many users experience, especially with fan-out

A good average can coexist with a painful tail that users regularly hit.

पाठ: Tail Latency and Fan-Out

Designing for Availability

13. What is split brain in an active-passive database cluster?

  1. Both nodes believing they are the primary and accepting writes
  2. Two replicas with different schemas
  3. A slow replica
  4. A load balancer with two algorithms
जवाब देखें

जवाब: Both nodes believing they are the primary and accepting writes

Split brain causes divergent writes; quorum-based leader election and fencing prevent it.

पाठ: Redundancy and Failover

14. The business can tolerate 4 hours of downtime and losing up to 24 hours of data for an internal tool. Which DR strategy is most appropriate?

  1. Multi-site active-active
  2. Warm standby in three regions
  3. No backups at all
  4. Backup and restore
जवाब देखें

जवाब: Backup and restore

Generous RTO and RPO targets are met cheaply by backup and restore.

पाठ: Zones, Regions and Disaster Recovery

15. Three components are each 99.9% available and all are required for a request. What is the approximate overall availability?

  1. 99.9%
  2. 99.97%
  3. About 99.7%
  4. 99.999%
जवाब देखें

जवाब: About 99.7%

0.999³ ≈ 0.997, so about 99.7%.

पाठ: Availability Arithmetic

Failure Modes and Defences

16. Why can a slow dependency be more dangerous than one that is completely down?

  1. Slow dependencies return wrong data
  2. Slow dependencies cannot be monitored
  3. Waiting requests pile up and exhaust threads and connections in every caller
  4. Load balancers route more traffic to slow services
जवाब देखें

जवाब: Waiting requests pile up and exhaust threads and connections in every caller

Hung calls hold resources, so the slowness spreads upstream.

पाठ: How Distributed Systems Fail

17. What does a circuit breaker do when a dependency keeps failing?

  1. Retries faster
  2. Stops calling it for a while and fails fast, then probes to see if it recovered
  3. Restarts the dependency
  4. Increases the timeout
जवाब देखें

जवाब: Stops calling it for a while and fails fast, then probes to see if it recovered

Opening the circuit protects both caller and dependency until a trial call succeeds.

पाठ: Timeouts, Retries, Circuit Breakers and Load Shedding

18. What is the main reliability benefit of a canary release?

  1. A bad change affects only a small share of traffic and is detected before full rollout
  2. It makes builds faster
  3. It removes the need for testing
  4. It doubles capacity
जवाब देखें

जवाब: A bad change affects only a small share of traffic and is detected before full rollout

Canaries limit blast radius and provide a comparison against the stable version.

पाठ: Safe Changes: Deployments, Flags and Rollbacks

Observability, Incidents and Testing

19. Which set lists the four golden signals?

  1. Latency, traffic, errors, saturation
  2. CPU, memory, disk, network
  3. Rate, errors, duration, uptime
  4. Throughput, cost, size, age
जवाब देखें

जवाब: Latency, traffic, errors, saturation

The SRE golden signals are latency, traffic, errors and saturation.

पाठ: Golden Signals and SLO-Based Alerting

20. During a major incident caused shortly after a deploy, what should usually happen first?

  1. Find the exact line of code at fault
  2. Write the postmortem
  3. Wait for the next scheduled deploy
  4. Mitigate, for example by rolling back the deploy, then investigate
जवाब देखें

जवाब: Mitigate, for example by rolling back the deploy, then investigate

Restoring service for users comes before full diagnosis.

पाठ: Incident Response and Postmortems

21. What is the purpose of a chaos engineering experiment?

  1. To randomly break production with no plan
  2. To replace monitoring
  3. To verify, with a hypothesis and limited blast radius, that the system tolerates a specific failure
  4. To measure developer productivity
जवाब देखें

जवाब: To verify, with a hypothesis and limited blast radius, that the system tolerates a specific failure

Chaos experiments test resilience assumptions deliberately and safely.

पाठ: Load Testing, Capacity Planning and Chaos Engineering

Putting It Into Practice and Revision

22. About how many requests per second is 864 million requests per day on average?

  1. 100
  2. 1,000
  3. 10,000
  4. 100,000
जवाब देखें

जवाब: 10,000

864,000,000 / 86,400 = 10,000 requests per second on average.

पाठ: Back-of-the-Envelope Capacity Estimation

23. Product pages are read thousands of times per minute and the primary database is busy. What is usually the first step?

  1. Shard the database by product ID
  2. Add caching (and a CDN) and serve reads from replicas
  3. Move to multi-region active-active
  4. Rewrite the app in another language
जवाब देखें

जवाब: Add caching (and a CDN) and serve reads from replicas

Read-heavy load is cheapest to address with caching and read replicas before partitioning.

पाठ: Case Study: Scaling a Web App Step by Step

24. Which item is part of a typical production readiness review?

  1. Tested backup restores with documented RPO and RTO
  2. Choosing a code formatter
  3. Picking the team logo
  4. Deciding variable names
जवाब देखें

जवाब: Tested backup restores with documented RPO and RTO

Readiness reviews focus on operability: SLOs, alerts, capacity, redundancy, backups and safe deploys.

पाठ: Production Readiness Review

25. Two independent replicas each have 99% availability, and either can serve requests. What is the combined availability?

  1. 98%
  2. 99%
  3. 99.5%
  4. 99.99%
जवाब देखें

जवाब: 99.99%

1 − (1 − 0.99)² = 1 − 0.0001 = 0.9999.

पाठ: Revision and Interview Questions

← कोर्स पर वापस जाएँ