# Amazon DynamoDB Accelerator (DAX) Cluster

> A DAX cluster bills per node-hour for each cache node from creation until deletion, with no stop operation and no per-request charge. ZopNight discovers DAX clusters every 6 hours, attributes cost per cluster from Cost Explorer or CUR 2.0, and flags caches fronting tables with little read traffic.

Source: https://zop.dev/integrations/aws/dax
Updated: 2026-08-19

---

DAX is an in-memory cache for DynamoDB, deployed as a cluster of cache nodes billed per node-hour. DAX clusters run continuously regardless of read traffic, so caches in front of quiet tables are a common source of forgotten spend.

## A flat rate for a read accelerator

DAX bills one way: per node-hour, for every node in the cluster, from creation until deletion. Requests through the cache are not charged. The cluster is the charge. That inverts DynamoDB's own economics, where on-demand tables cost nothing at rest: put a DAX cluster in front of one and the pair now has a fixed hourly floor no matter how little the application reads.

## A cache must earn its keep

DAX pays for itself in exactly one situation: read-heavy, hot-key traffic where cache hits absorb reads that would otherwise consume table capacity, or where microsecond latency is a requirement. Outside that shape it is pure overhead. The break-even is measurable, since the node-hours must cost less than the read capacity they save, and a cache fronting a table with modest traffic never crosses it.

## How DAX spend goes stale

The launch-day cache: added while chasing a latency target, kept after the feature quietly changed. The multi-node dev cluster: three nodes of high-availability caching for a test table nobody reads. And clusters surviving their tables: the application moved on, the table traffic died, the cache nodes kept billing hourly.

## Tracking the cluster, not the requests

DAX clusters are discovered on the 6-hour cycle, with per-cluster cost attributed from Cost Explorer or CUR 2.0 and recommendations flagging idle clusters, meaning caches whose fronted tables show little read traffic. There is nothing to schedule: like ElastiCache, the choice is running or deleted.

## Cluster inventory in the console

DynamoDB console, then DAX, then Clusters in the left navigation. Node count and node type per cluster sit on the cluster page; the fronted table's read metrics, next door under Tables, tell you whether the cache is earning its node-hours.