Google Cloud Memorystore Cost: Redis and Valkey Pricing

Memorystore is billed from the capacity you provision, even when the cache is idle or stores less data. For Memorystore for Redis, the service tier, capacity tier, provisioned GiB, and region set the instance charge. Basic Tier is a standalone cache; Standard Tier adds cross-zone replication and automatic failover, so it costs more for the same primary capacity.

Start with the working set, then reserve headroom for expiration delays, replication, fragmentation, and traffic spikes. Multiply the applicable regional hourly rate by provisioned capacity and monthly running hours. Capacity tiers can change the per-GiB rate and network performance, so compare the exact tier before rounding the estimate up.

Memorystore for Valkey uses node-based pricing. Estimate the number of shards, add every replica node, and multiply the total by the selected node type's regional hourly price. Replicas improve availability and read throughput, but each replica is another provisioned node rather than free redundancy.

Add backup storage and AOF persistence when enabled. Backups have a minimum 24-hour charge and are not automatically deleted with the instance, while AOF is charged from provisioned instance size. Keep separate budget lines for compute nodes, persistence, and retained backups so old recovery points do not become an invisible recurring cost.

A cache should reduce a more expensive bottleneck. Compare the monthly Memorystore bill with database reads, queries, CPU time, and latency it actually avoids. Track memory utilization, evictions, hit ratio, command latency, and the source database load; a large idle cache or a low hit ratio is a signal to resize, change expiration policy, or remove the cache.

Place clients in the same region and, where practical, the same zone. Redis access in the same region is not billed as Memorystore ingress or egress, but client resources can still incur cross-zone charges; cross-region access adds network cost and latency. Model normal traffic and a cache-miss surge before choosing a tier.

Estimate the Cloud Run and Cloud SQL costs a cache may reduce.