Exercises › Distributed Systems Consensus › CAP Theorem and PACELC Discussion

CAP Theorem and PACELC Discussion

5 questions · Distributed Systems Consensus

1. An interviewer asks: 'If your system experiences a network partition, do you prioritise consistency or availability?' You answer that your database returns an error if it cannot confirm consistency rather than serving potentially stale data. What trade-off does this describe?

AP: prioritising Availability over Consistency CP: prioritising Consistency over Availability CA: prioritising Consistency and Availability

2. The PACELC theorem extends CAP by adding which consideration that CAP ignores?

Partition recovery time Latency vs. Consistency trade-off even without partitions The cost of replication

3. A DynamoDB table is configured for eventual consistency. After a write to one region, reads from a different region may return the old value for up to 1 second. This is an example of:

A quorum read delay Eventual consistency replication lag Causal inconsistency

4. During a design review, a senior engineer argues that 'you can never have both CA in a distributed system with network partitions.' Why is this correct?

Because distributed systems always have low latency Because network partitions are inevitable, so the P in CAP cannot be avoided Because consistency and availability are always conflicting goals

5. A team is debating between Cassandra (AP) and CockroachDB (CP) for their global e-commerce product catalogue. Which scenario would best justify choosing the AP database?

The catalogue must show the exact same price to all users at all times Temporary stale reads are acceptable and high availability is critical Inventory counts must be immediately consistent across regions

Exercise complete!

out of 5 questions

Back to Distributed Systems Consensus All exercises