#05Distributed Cache Cluster
5 / 20
Amazon logoAmazon System
MediumDistributed Systems
30–45 min

Distributed Cache Cluster

Design a distributed in-memory key-value caching system supporting consistent hashing, replication, write-through/write-back policies, and LRU eviction.

Scale & Traffic1M operations/sec · 500GB cache cluster
Monthly Cloud Budget$4,000/month
Latency Target (P99)< 50ms Realtime SLA
Availability SLA99.99% Multi-AZ Active

System Requirements

  • Sub-millisecond read/write latency at 99.9th percentile
  • Consistent hashing with virtual nodes to minimize rebalancing keys
  • Master-replica redundancy with automated failover
  • Configurable eviction policies: LRU, LFU, and TTL expiration

Graded Constraints

01Cost limit: $4,000/month
02Multiple cache nodes (Redis / Memcached) partitioned logically
03Load Balancer or Proxy for client key routing
04Persistent backing database for cache misses

Ready to design?

Open this challenge in the interactive system design playground and satisfy all architecture constraints.

Explore More

Related Architecture Challenges

All Problems
Distributed Cache Cluster — System Design Problem | Mimir Nest