ElastiCache Serverless

With the engine decision behind you—whether Valkey, Redis OSS, or Memcached—the next architectural choice determines how that engine actually runs in production. AWS offers two distinct deployment models for ElastiCache, and the one you pick reshapes every operational conversation that follows, from provisioning and scaling to patching and client connectivity. This lesson focuses exclusively on ElastiCache Serverless, the model where AWS owns the infrastructure shape and your application interacts with a single, stable DNS endpoint. The core promise is straightforward: you never select a node type, never decide how many shards to create, and never schedule a maintenance window for engine patches. Behind that simplicity sits a managed proxy layer, an automatic scaling mechanism, and a consumption-based pricing model that together change how teams think about caching infrastructure. The sections ahead cover capacity abstraction, the proxy layer and single endpoint, operational simplicity in day-2 tasks, and the cost trade-offs that determine whether serverless is the right fit for a given workload. The following lesson will then examine the alternative node-based cluster model, so the comparison stays clean.

What capacity abstraction means

In the traditional ElastiCache deployment model, a team makes several upfront decisions before a single key is written. They choose an instance family such as r6g or r7g, pick a specific size like large or xlarge, set the number of shards for horizontal partitioning, and decide how many read replicas each shard should carry. After launch, they monitor CPU utilization, memory pressure, and network throughput through CloudWatch, and when thresholds are breached, they trigger a manual or policy-driven resize that can involve downtime risk or connection churn.

ElastiCache Serverless removes every one of those decisions from the deployment workflow. When you create a serverless cache, you specify the engine, a name, security configuration, and VPC placement. AWS handles everything else.

Note: Capacity abstraction does not mean capacity is unlimited. AWS scales resources behind the scenes, but the system still operates within service quotas and the physical constraints of the underlying infrastructure.

The serverless model measures usage along two independent dimensions. ...