Scaling & Distributed Databases Quiz

Q1. In a Dynamo-style system with N = 3 replicas, which R/W quorum pair guarantees that every read sees the latest successful write? R = 1, W = 1
R = 2, W = 2
R = 1, W = 2
R = 2, W = 1

Q2. An application is deployed across multiple data centers and must continue accepting writes even if network connectivity between sites is lost. Which replication strategy is most appropriate for this scenario? Single-leader (primary-follower) replication
Multi-leader (active-active) replication
Leaderless quorum-based replication
Vertical scaling on a single server

Q3. In a primary-secondary database setup using asynchronous replication, what is a notable drawback compared to synchronous replication? Write latency significantly increases on the primary
The primary cannot serve reads while replicating
A crash of the primary may lead to recent writes not being replicated to secondaries
Secondary nodes cannot serve read traffic at all in this mode

Q4. You need to shard a time-series database by timestamp. Which sharding approach makes range queries (by time) efficient but risks hotspots if the data is unevenly distributed? Hash-based sharding on the timestamp
Consistent hashing on the record ID
Random sharding of records to nodes
Range-based sharding by timestamp values

Q5. What is the main benefit of using consistent hashing to distribute keys across nodes? It guarantees perfect balance of load across all nodes at all times
It eliminates the need for replication of data
When nodes are added or removed, only a minimal portion of the data needs to move
It ensures all writes are strongly consistent without coordination

Q6. If the primary node of a database cluster fails, which mechanism helps the remaining nodes agree on electing a new primary leader? Running a consensus algorithm (like Paxos or Raft) among the nodes
Initiating a two-phase commit (2PC) across all replica nodes
Using vector clock timestamps to pick the latest node
Relying on manual failover procedures only

Q7. In a microservices distributed transaction, what is an advantage of using the Saga pattern over two-phase commit (2PC)? A Saga allows each service to commit locally and uses compensating transactions on failure, avoiding a global lock/blocking on one coordinator
A Saga guarantees strict serializable consistency across all services just like 2PC
The Saga pattern uses a single coordinator to enforce atomic commits across services
The Saga pattern will roll back all services immediately on any failure, rather than doing partial commits

Q8. Google Spanner can perform distributed transactions across data centers with strong consistency. What technology does Spanner use to achieve externally-consistent timestamps globally? Two-phase commit protocol with retries
Vector clocks to order transactions
The TrueTime API based on atomic clocks and GPS clocks
A logical clock derived purely from message ordering

Q9. In Amazon’s Dynamo system, vector clocks are used to: Synchronize physical clocks of all nodes to the same time
Detect conflicting concurrent updates by capturing causality of updates
Elect a leader among replicas during failures
Optimize the two-phase commit process

Q10. Which statement about Conflict-Free Replicated Data Types (CRDTs) is correct? They rely on a central coordinator to resolve conflicts
They require all updates to go through a single leader
They only work if all updates are applied in the same order on every node
They allow concurrent updates on multiple nodes to merge automatically into a consistent state without conflicts

Q11. According to the CAP theorem, an “AP” distributed database (Available and Partition-tolerant) will sacrifice which property during a network partition? Availability
Partition tolerance
Consistency
Durability

Q12. In the two-phase commit (2PC) protocol, if the coordinator crashes after sending the “prepare to commit” messages (and all participants replied ready) but before sending a commit/abort, what happens to the participant nodes? They automatically commit since they voted to commit
They automatically abort the transaction to avoid inconsistency
They elect a new coordinator to decide commit or abort immediately
They wait blocked, holding the transaction in doubt until the coordinator (or a recovery process) resolves it

Q13. One database shard has become a hotspot because a single key is receiving an extremely high volume of traffic. Which strategy helps mitigate this hot-key issue? Reducing the total number of shards to concentrate capacity
Splitting or salting the key (e.g., add a random prefix/hash) to distribute data across multiple shards
Disabling replication for that particular key’s data
Using a single-threaded server to serialize all operations on that key

Q14. Which guarantee is provided by an eventually consistent database system? Every read receives the very latest write value
No need for any conflict resolution, as conflicts cannot occur
Writes are immediately visible on all replicas
If no new updates are made to an item, all replicas will eventually have the same last updated value for that item

Q15. In a distributed database cluster, a “split-brain” scenario refers to: A network partition splitting the cluster into two groups that each think they are the primary cluster, leading to conflicting state
A situation where the database’s memory is exhausted and it can no longer process requests
A design where the database is sharded into two halves for scaling
A backup system that is out of sync with the primary database

system-design