We optimized the key-value store by reducing the number of load balancers, caches, and database instances. We also replaced 70 small application servers with four larger instances, achieving better cost efficiency while preserving scalability and fault tolerance. The final design met all latency, durability, and capacity requirements with a lower cost than the reference architecture.
Sign in to join the discussion.
fewer boxes, same grade. nice
did the math on the consolidation. 70 app.small at $0.04 is $2.80/hr for 35k req/s of capacity, 4 xlarge is $2.72 for 32k. basically a wash on paper except you also drop the lb fanout and deploys take a tenth as long. good trade