Ensure Scalability and Replication

Add scalability

Scalability requires distributing data across multiple storage nodes. As demand changes, we must dynamically add or remove nodes. To achieve this, we partition data to balance the load across the system.

A simple placement rule hashes a key and takes its remainder modulo m, the number of nodes. The result x selects a node. With four nodes and a suitable hash, this can distribute key counts roughly evenly; it does not imply that each node receives exactly 25% of requests.

The following slides explain this process: