Size the disk your cluster needs with the published AWS sizing formula, then price it with list prices you can see and edit. Defaults are us-east-1 On-Demand prices as of October 2026; verify current prices before you budget.
1,362 GB
454 GB per data node across 3 nodes, from 500 GB of source data
$550
Excludes dedicated master or coordinating nodes, extra provisioned IOPS or throughput, snapshots, data transfer, support and tax. On-Demand list prices; your discounts are not applied.
Storing embeddings? Size vector memory with the vector sizing calculator.
Storage follows the AWS OpenSearch Service sizing guide (Calculating storage requirements):
source x (1 + replicas) x index factor / (1 - 0.05) / (1 - headroom)
With 1 replica, a 1.1 index factor (AWS: index plus source is often 110% of source) and the 20% OpenSearch Service overhead, this is AWS's worked example: 66 x 2 x 1.1 / 0.95 / 0.8 = 191 GiB, which AWS simplifies to source x (1 + replicas) x 1.45. The 5% is Linux's default root-reserved space. For self-managed clusters and Elastic Cloud the headroom defaults to 15%, because Elasticsearch's default low disk watermark stops shard allocation above 85% disk used (disk-based shard allocation). Rolling data uses ingest per day x retention days as the source size, as the AWS guide describes.
Cost = data nodes x price per node-hour x 730 hours (24 x 365 / 12) + storage required x price per GB-month. For Elastic Cloud Hosted, cost = data nodes x RAM per node x price per GB RAM-hour x 730; Elastic bills deployments by size, with storage bundled in the hardware profile (Elastic Cloud pricing). Check that the profile you pick gives each node at least the per-node storage shown.
Default prices are us-east-1 On-Demand list prices from AWS's published price list as of October 2026: EC2 (Linux), EBS and OpenSearch Service. Elastic Cloud prices depend on region, tier and hardware profile, so enter yours. Prices change; verify before you budget.
What you save depends on your data and queries, so we don't put a number on it. These are the levers the vendors document.
Each replica is a full copy, so it multiplies storage. AWS recommends at least one to prevent data loss; indices you can reindex from source may not need more (AWS sizing guide).
Move older indices to warm, cold or frozen tiers with ILM (Elastic data tiers), or ISM with UltraWarm on OpenSearch (ISM, UltraWarm). UltraWarm managed storage lists at $0.024 per GB-month in us-east-1 versus $0.122 for gp3.
index.codec: best_compression uses ZSTD; Elastic reports up to ~28% lower storage on log datasets, with slower get-by-id (index settings).
Compare instance families in the price list (r6g vs r6i above), and price Reserved Instances (OpenSearch Service) or EC2 Savings Plans for steady clusters.
AWS sizing guidance: source data x (1 + replicas) x (1 + indexing overhead) / (1 - 0.05 Linux reserved) / (1 - 0.2 OpenSearch Service overhead), simplified to source x (1 + replicas) x 1.45. For 100 GB of source data with one replica, that is about 290 GB.
From AWS published us-east-1 On-Demand list prices as of October 2026 for EC2 (Linux), EBS and Amazon OpenSearch Service. Every price is editable, and Elastic Cloud prices must be entered from Elastic's own pricing calculator. Verify current prices before budgeting.
Dedicated master and coordinating nodes, extra provisioned IOPS or throughput, snapshots, data transfer, support, tax and any discounts. It prices data nodes and their storage only.
The documented levers are fewer replicas for indices you can rebuild, tiering older data with ILM or ISM and UltraWarm, best_compression for stored fields, Graviton instance families, and Reserved Instances or Savings Plans for steady clusters. How much each saves depends on your data and queries.
We look at your mappings, shards, replicas and tiers, and size the cluster to your actual workload.
