Exercises › Distributed Systems Consensus › Distributed Clocks and Ordering

Distributed Clocks and Ordering

5 questions · Distributed Systems Consensus

1. Each server independently assigns a timestamp to events using its local clock. Server A's clock is slightly ahead. An event on Server B at time T1 and an event on Server A at T1+1ms may actually have occurred in the opposite order. What problem does this illustrate?

Clock synchronisation overhead Clock skew and the limits of physical clocks in distributed systems Event ordering bias

2. A counter that increments with each event on a node, and when a message is received, updates to max(local, received) + 1. This allows ordering events that have a causal relationship. What is this called?

A vector clock A Lamport timestamp A hybrid logical clock

3. To determine if event A on Server 1 happened before event B on Server 2 (not just the Lamport order), you need to track causality per-node. What data structure is used?

A Lamport timestamp A vector clock A timestamp oracle

4. A database needs to assign globally monotonically increasing timestamps with wall-clock accuracy for distributed transactions, supporting both logical ordering and real-time ordering. What type of clock does Spanner use?

A vector clock A TrueTime / hybrid logical clock (HLC) A Lamport timestamp

5. When designing a distributed system, an architect says: 'We cannot rely on timestamps to determine event order — instead we use explicit causal metadata in every message.' Which approach are they describing?

Timestamp-based linearisability Causal tracking via message metadata (vector clocks or causal tokens) Clock synchronisation via NTP

Exercise complete!

out of 5 questions

Back to Distributed Systems Consensus All exercises