Data nodes plus EBS storage for a managed search cluster.
Search clusters get expensive when stale indices force more or bigger hot nodes. Roll old data to warm/cold tiers and size hot nodes to live queries. Related: log retention calculator.
The OpenSearch / Elasticsearch Cost Calculator estimates the monthly bill for running a search cluster's data tier. It takes the number of data nodes, the price per node per hour, and the hours per month the cluster runs, then multiplies them to produce your recurring compute cost. The EBS storage per node input scales that total across the cluster, so you can see how much attached disk your nodes carry alongside the instances themselves. The main drivers are simple and direct: adding nodes, choosing a larger (more expensive) hourly instance, or running around the clock all push the figure up in a straight, predictable line.
The key trade-off to watch is node count and instance size versus storage per node. A cluster with many small nodes and heavy per-node storage can cost far more than a leaner design that holds the same data, so it pays to right-size before committing. Because hours per month is an explicit input, the calculator also makes the cost of always-on clusters visible — non-production environments that idle overnight or on weekends rarely need a full month of runtime. Enter the real numbers from your provider's pricing page rather than round estimates, and re-run it whenever you scale nodes or storage to keep the projection honest.
Per data-node-hour for compute, plus EBS storage per GB-month attached to each node, plus dedicated master nodes on larger clusters. You pay for the running cluster continuously, not per query.
Use index lifecycle management to move or delete old indices, keep only the shards you query hot, and right-size node memory to your searchable data. UltraWarm and cold storage tiers hold rarely-searched data far cheaper than hot nodes.