AWS OpenSearch Service (successor to Amazon Elasticsearch Service) offers managed search and analytics. You can choose between managed clusters with full control or OpenSearch Serverless for automatic scaling. See the OpenSearch pricing page for current rates. The pricing difference between the two is substantial — and picking the wrong option for your workload can triple your costs.
TL;DR: Managed OpenSearch clusters start at $0.036/hour for t3.small.search (about $26/month). Serverless costs $0.24/OCU-hour with a minimum of 4 OCUs ($700/month minimum). For small search workloads, managed clusters are dramatically cheaper. Serverless is best for unpredictable large-scale workloads where you want zero capacity management.
Managed Cluster Pricing
Popular Instance Types
| Instance | vCPUs | RAM | On-Demand/hr | Monthly |
|---|---|---|---|---|
| t3.small.search | 2 | 2 GB | $0.036 | $26 |
| t3.medium.search | 2 | 4 GB | $0.073 | $53 |
| m6g.large.search | 2 | 8 GB | $0.128 | $93 |
| r6g.large.search | 2 | 16 GB | $0.167 | $122 |
| r6g.xlarge.search | 4 | 32 GB | $0.335 | $245 |
| r6g.4xlarge.search | 16 | 128 GB | $1.340 | $978 |
Storage Costs
| Storage Type | Cost |
|---|---|
| GP3 (General Purpose SSD) | $0.122/GB/month |
| GP2 (older SSD) | $0.135/GB/month |
| IO1 (Provisioned IOPS) | $0.138/GB/month + $0.10/IOPS |
| Magnetic (previous gen) | $0.084/GB/month |
UltraWarm (Warm Storage)
For infrequently accessed data at lower cost:
| Component | Cost |
|---|---|
| ultrawarm1.medium | $0.238/hr ($174/mo) — 1.5 TB |
| ultrawarm1.large | $0.476/hr ($348/mo) — 20 TB |
UltraWarm is approximately 90% cheaper per GB than standard hot storage for log analytics and time-series data.
OpenSearch Serverless Pricing
| Component | Cost |
|---|---|
| Indexing compute | $0.24/OCU-hour |
| Search compute | $0.24/OCU-hour |
| Managed storage | $0.024/GB/month |
| Minimum OCUs | 2 indexing + 2 search = 4 OCUs |
Minimum monthly cost: 4 OCUs x 730 hours x $0.24 = $701/month — even with zero traffic.
Managed vs Serverless Cost Comparison
| Scenario | Managed Cluster | Serverless |
|---|---|---|
| Small search (50 GB, low traffic) | $175/mo (2x m6g.large + GP3) | $701/mo (minimum) |
| Medium search (500 GB, moderate) | $550/mo (3x r6g.large + GP3) | $750-1,200/mo |
| Large analytics (5 TB, heavy) | $3,500/mo (multi-node r6g) | $2,800-4,000/mo |
| Unpredictable bursts | Sized for peak: $3,500/mo | $1,500-3,000/mo (auto-scales) |
Clear winner: Managed clusters for predictable workloads under $700/month. Serverless for large, unpredictable workloads.
Hidden Costs
Multi-AZ Deployment
Production clusters should span 3 AZs with dedicated master nodes. This means:
- 3 data nodes minimum (3x the single-node cost)
- 3 dedicated master nodes (additional cost)
- Cross-AZ data transfer at $0.01/GB
Dedicated Master Nodes
Required for production clusters with more than 10 nodes:
| Master Node | Cost/hr |
|---|---|
| m6g.large.search | $0.128 |
| r6g.large.search | $0.167 |
Three master nodes for high availability add $280-365/month.
Custom Endpoints and VPC
VPC access is free, but PrivateLink endpoints add charges if connecting from other VPCs. Custom domains (CNAME) are included at no extra cost.
Cost Optimization Strategies
1. Use UltraWarm for Log Retention
Move logs older than 7-14 days from hot to UltraWarm storage. Refer to the OpenSearch developer guide for configuration details. For 10 TB of logs, UltraWarm costs ~$350/month vs ~$1,220/month on GP3.
2. Apply Index Lifecycle Policies
Automate data movement: hot (7 days) → UltraWarm (30 days) → cold (S3, 90 days) → delete. This reduces storage costs by 70-80%.
3. Use Graviton Instances
m6g and r6g (Graviton) instances are 10-15% cheaper than equivalent x86 (m5, r5) instances with comparable performance.
4. Reserve Instances for Stable Workloads
1-year RIs save approximately 30%. 3-year RIs save approximately 50%. Only reserve your baseline capacity — handle peaks with on-demand nodes.
5. Optimize Shard Count
Over-sharding wastes memory and compute. Use 30-50 GB per shard as a target. Fewer, larger shards reduce overhead on cluster management.
Related Guides
- AWS ElastiCache Pricing Guide
- AWS Redshift Pricing Guide
- AWS DynamoDB Pricing Guide
- AWS Bedrock Embeddings Guide
FAQ
When should I use OpenSearch Serverless vs managed?
Use managed clusters when your monthly cost would be under $700 or when you need fine-grained control over instance types and configurations. Use Serverless when you want zero management and your workload justifies the $700+/month minimum.
How does OpenSearch pricing compare to Elastic Cloud?
AWS OpenSearch managed clusters are typically 10-20% cheaper than Elastic Cloud for equivalent configurations. However, Elastic Cloud includes features like machine learning and advanced security at no extra tier cost.
Can I reduce OpenSearch costs without losing search performance?
Yes. Use UltraWarm for older data, optimize shard sizing, enable GP3 instead of GP2 storage, and right-size instance types based on actual CPU and memory utilization from CloudWatch.
Lower Your OpenSearch Costs with Wring
Wring helps you access AWS credits and volume discounts to lower your OpenSearch costs. Through group buying power, Wring negotiates better rates so you pay less per node hour.
