Practice English vocabulary for distributed systems concepts: consensus, fault tolerance, partitioning, replication, and distributed transactions.
0 / 8 completed
1 / 8
What does 'fault tolerance' mean in a distributed system?
Fault tolerance means the system keeps working despite partial failures. Techniques include replication, circuit breakers, and fallback strategies.
2 / 8
What is 'split-brain' in the context of distributed systems?
Split-brain occurs when a network partition isolates nodes that each believe they are the sole active leader, potentially causing conflicting writes or inconsistent state.
3 / 8
What is 'quorum' in distributed consensus?
A quorum is the minimum number of nodes that must agree (vote) before a distributed operation is considered committed. For a 5-node cluster, a quorum is typically 3 (majority).
4 / 8
How would you explain 'vector clock' to a colleague?
A vector clock assigns a logical timestamp per node to each event. By comparing vector clocks, systems can determine causality (A happened before B) or detect concurrent events that may conflict.
5 / 8
What is 'gossip protocol' in distributed systems?
Gossip (epidemic) protocols spread information through random peer-to-peer exchanges. They are resilient, scalable, and eventually consistent — used by Cassandra, Redis Cluster, and Consul for cluster membership.
6 / 8
What does 'two-phase commit (2PC)' guarantee?
2PC uses a coordinator to prepare all participants (phase 1), then commit (phase 2) only if all participants are ready. It ensures atomicity across nodes but is blocking and not partition-tolerant.
7 / 8
What is 'backpressure' and when do engineers discuss it?
Backpressure prevents fast producers from overwhelming slow consumers. When a consumer is at capacity, it signals the producer to slow down. Without backpressure, queues grow unbounded and systems crash.
8 / 8
What does 'shard key' determine in a sharded database?
The shard key determines how data is distributed across shards. Using user_id as a shard key distributes writes evenly; using a timestamp can create hot spots where all writes go to the same shard.