Practical guide
ElastiCache pricing: count primaries, replicas, and node-hours
estimate ElastiCache Redis OSS node and replica costs. Includes a worked example, calculator scope, and practical questions.
Updated · Sources checked
Count every provisioned node in the topology
For node-based ElastiCache, the compute estimate follows billable node-hours. A replica is a running node, not a free annotation attached to the primary. Shard count and replicas per shard can therefore materially change the subtotal.
The calculator covers Redis OSS cache.t3 On-Demand nodes in US East. It does not estimate Valkey, Memcached, serverless usage, every node family, or a complete high-availability architecture.
Worked example: three shards and one replica each
Assume three shards, each containing one primary and one replica. The topology has six nodes. At 730 hours per node, total usage is 4,380 node-hours.
If the selected nodes all have hourly rate R, the modeled subtotal is 4,380R. Enter six nodes at 730 hours. Entering three would count only the primaries and omit half of this example’s node compute.
Now consider two replicas per shard. Total node count becomes nine and usage rises to 6,570 node-hours. The additional replica layer adds 2,190 node-hours, a 50% increase over the six-node topology. This arithmetic does not prescribe a replication policy; availability and read requirements determine whether that topology is appropriate.
Match the estimate to a real configuration
Confirm that the chosen engine, node type, and topology are supported for your deployment. Count temporary overlap during migrations or scaling if old and new nodes are simultaneously billable. A steady-state estimate can understate a changeover month.
The tool models standard node usage only. Backup, transfer, reserved-node terms, serverless metering, and other excluded charges require separate review. It also does not calculate usable cache capacity after application data structures and memory overhead.
Track hit rate, memory pressure, evictions, connections, and latency before changing the number or size of nodes. A smaller node bill is not a useful outcome if extra cache misses overload the backing database.
Frequently asked questions
Are replicas included automatically? Include their nodes explicitly in the count; the calculator does not discover the topology.
Can I use this for ElastiCache Serverless? No. Serverless uses a different pricing model; use the applicable official service dimensions rather than substituting a guessed node count.
Calculate your scenario
Use the elasticache cost calculator. Keep the displayed region, pricing date, billing units, and exclusions alongside your result. The arithmetic examples above illustrate usage or explicitly hypothetical rates; they are not AWS quotes.
Source
Billing structure checked against official AWS documentation on September 6, 2026.