What best characterizes a hot shard anti-pattern?

What best characterizes a hot shard anti-pattern?

• A. A read-heavy system with no writes.
• B. All shards receiving equal load due to perfect distribution.
• C. A shard receiving disproportionately heavy traffic becoming a bottleneck.
• D. A multi-region replication strategy that slows writes.

Answer: (C)

Hot shard anti-pattern shows up when the distribution of work across shards is uneven. In a sharded system you want requests and data to be spread out, but if one shard ends up handling a large share of the traffic—often because the shard key causes many accesses to map to that same shard—that shard becomes a bottleneck. The system’s overall performance is then limited by that single shard’s capacity, even if other shards are underutilized. That’s exactly what the option describes: a shard receiving disproportionately heavy traffic becoming a bottleneck. The other scenarios don’t capture this imbalance across shards: a read-heavy setup with no writes isn’t inherently about a single shard bottleneck, perfectly balanced distribution means no hot shard, and replication that slows writes relates to cross-region latency rather than shard-level load concentration.

Hot shard anti-pattern shows up when the distribution of work across shards is uneven. In a sharded system you want requests and data to be spread out, but if one shard ends up handling a large share of the traffic—often because the shard key causes many accesses to map to that same shard—that shard becomes a bottleneck. The system’s overall performance is then limited by that single shard’s capacity, even if other shards are underutilized. That’s exactly what the option describes: a shard receiving disproportionately heavy traffic becoming a bottleneck. The other scenarios don’t capture this imbalance across shards: a read-heavy setup with no writes isn’t inherently about a single shard bottleneck, perfectly balanced distribution means no hot shard, and replication that slows writes relates to cross-region latency rather than shard-level load concentration.