Work out cache node sizing instantly with clear inputs, formula shown and shareable results.
Cache sizing reserves memory for the share of the working set you want resident, then divides by usable memory per node — around 75 percent of nominal, because Redis needs headroom for replication buffers, fragmentation and background saves. Replicas multiply node count without adding cache capacity, since they hold the same data.
Cache sizing
memory needed = working set x target hit rate; usable per node = node memory x 0.75; shards = ceil(needed / usable); nodes = shards x (1 + replicas)
Redis needs room for the replication backlog, copy-on-write during background saves, and allocator fragmentation. Filling it causes eviction storms or OOM.
No, they hold identical data for availability and read scaling. Adding capacity requires more shards, not more replicas.