Consensus Discussion Language

1. During an architecture review an engineer says: 'We need to handle network partitions — our service spans three availability zones.' Which statement correctly explains why this consideration matters for a consensus-based system?

Network partitions only occur in on-premise data centres, not cloud AZs A network partition between availability zones can prevent nodes from communicating — a consensus protocol must decide whether to sacrifice availability or consistency during such a partition Network partitions only affect storage, not computation

2. A monitoring runbook states: 'Leader election takes approximately 150ms in this cluster.' What does this figure represent in terms of system behaviour?

The time for the new leader to replicate all existing log entries to followers The window of unavailability between a leader failing and a new leader being elected and ready to serve requests The network round-trip time between nodes in the cluster

3. In an architecture review a team member says: 'The quorum is 3 of 5 nodes.' Which statement correctly interprets this and its implication for fault tolerance?

The cluster requires all 5 nodes to be available to commit a write — if any node fails, writes are rejected A majority of 3 nodes must agree for elections and commits — the cluster can tolerate up to 2 simultaneous node failures and remain operational The quorum of 3 means only 3 nodes store data; the other 2 are read-only replicas

4. A team is debating between a 3-node and a 5-node Raft cluster. One engineer argues: 'A 5-node cluster tolerates 2 failures versus 1 for a 3-node cluster, but every write requires 3 acknowledgements instead of 2, adding latency.' How would you describe this discussion?

This is a discussion about storage cost, not a design trade-off This is a consensus design trade-off: more nodes improve fault tolerance but increase coordination overhead and write latency — a classic availability vs. performance discussion This is a security trade-off: more nodes increase attack surface

5. In a design review, an engineer proposes using a consensus protocol and says: 'For our use case, we can accept slightly higher write latency in exchange for strong consistency guarantees.' Which phrasing correctly summarises this trade-off position?

'We sacrifice consistency to achieve low latency' 'We prioritise correctness over write throughput — the consensus overhead is acceptable for our workload' 'We need eventual consistency because strong consistency is not achievable'

Vocabulary Reference

Phrase	Usage in architecture discussions
we need to handle network partitions	Signals that the design must explicitly address the CAP trade-off — the system spans nodes that can lose connectivity.
leader election takes ~150ms	Quantifies the write unavailability window after a leader failure; used to discuss SLA impact of a consensus-based system.
the quorum is 3 of 5 nodes	States the minimum number of nodes that must agree for operations to proceed and implies the fault tolerance: up to 2 node failures can be survived.
we sacrifice availability during a partition	Describes a CP design choice — the system prefers to return errors rather than serve potentially stale data.
consensus overhead	The extra latency and coordination round-trips required by a consensus protocol; the cost paid for strong consistency.
design trade-off language	In architecture reviews: name what you gain ("strong consistency", "fault tolerance"), what you accept ("higher latency", "reduced availability"), and why that is acceptable for your workload.

Vocabulary Reference

Exercise complete!