Amazon OpenSearch Service Pricing

Except as otherwise noted, our prices are exclusive of applicable taxes and duties, including VAT and applicable sales tax. For customers with a Japanese billing address, use of AWS is subject to Japanese Consumption Tax. Learn more.

With Amazon OpenSearch Service Reserved Instances, you can reserve instances for a one- or three-year term and realize significant savings on usage costs compared to On-Demand instances. Functionally, On-Demand and Reserved Instances are identical. From a billing perspective, however, Reserved Instances can provide significant cost savings.

Reserved Instances have three payment options:

• No Upfront Reserved Instances (NURI) – NURIs offer significant savings compared to On-Demand Instance pricing. You pay nothing upfront, but you commit to pay for the Reserved Instances over the course of a one- or three-year term. One-year NURIs offer a 31% discount and three-year NURIs offer a 48% discount. For T3.medium, one-year NURIs offer a 18% discount and three-year NURIs offer a 28% discount.
  

• Partial Upfront Reserved Instances (PURI) – PURIs offer higher savings than NURIs. This option requires you to pay a portion of the total cost upfront and pay the remainder of the cost on an hourly basis over the course of the term. One-year PURIs offer a 33% discount and three-year PURIs offer a 50% discount. For T3.medium, one-year PURIs offer a 20% discount and three-year NURIs offer a 30% discount.
  

• All Upfront Reserved Instances (AURI) – AURIs offer the highest savings of all of the Reserved Instance payment options. You pay for the entire reservation with one upfront payment and pay nothing on an hourly basis. One-year AURIs offer a 35% discount and three-year AURIs offer a 52% discount. For T3.medium, one-year AURIs offer a 22% discount and three-year NURIs offer a 32% discount.

• Reserved Instance pricing is specific to each region and depends on the payment option and term that you select. When you purchase a Reserved Instance, you will be charged the associated upfront fees (if applicable) and hourly fees (if applicable), even if you are not currently running Amazon OpenSearch Service. To purchase Reserved Instances, visit the Reserved Instance tab in our Console.
  

Your monthly charge is based on how long your instance runs that month:

• Hourly-billed instances: Hours in the month × hourly usage rate

The hourly usage rate equals the total Reserved Instance cost divided by the total hours in the term (calculated using a 365-day year)
 

With Amazon OpenSearch Serverless, you only pay for the resources consumed by the workload. OpenSearch Serverless charges for compute and storage separately. The compute capacity is measured in OpenSearch Compute Units (OCUs). The number of OCUs corresponds directly to the CPU, memory, Amazon EBS storage, and I/O resources required to index data or run queries. One OCU comprises 6 GB of RAM, corresponding vCPU, GP3 storage (used to provide fast access to the most frequently accessed data), and data transfer to Amazon Simple Storage Service (S3).

You will see one entry for compute in OCU-hours with two labels: one for data indexing and the other for search. OCUs are billed on an hourly basis on a collection with per-second granularity. Data stored on Amazon S3 will be billed by gigabyte-months. You will be billed at least for a minimum of 2 OCUs (1 OCU [0.5 x 2] indexing includes primary and standby, and 1 OCU [0.5 x 2] search includes one replica for HA) for the first collection in an account. However, the minimum varies based on your data size and collections type in use.

Additionally, OpenSearch Serverless also offers a dev-test option, where you can launch a collection without redundant standby nodes. This deployment mode further cuts the cost in half, with 0.5 OCU for indexing and 0.5 OCU for search. All the data is stored in Amazon S3 in both modes offering the complete data durability. However, the minimum varies based on your data size and collections type in use.

All subsequent collections using the same encryption key can share those OCUs. Additional OCUs will be added based on compute instances and data needed to support your collections. You can configure a maximum number of OCUs per account to control costs.

A vector search collection cannot share OCUs with search and time series collections, even if the vector search collection uses the same KMS key as the search or time series collections. A new set of OCUs will be created for your first vector search collection. The OCUs of vector search collections are shared among the same KMS key vector collections.

With Amazon OpenSearch Ingestion, you only pay for the resources consumed by your workload. OpenSearch Ingestion charges for only the compute needed to ingest, transform, and route data in an OpenSearch Ingestion pipeline. The compute capacity is measured in OpenSearch Compute Units (OCUs). The number of OCUs corresponds directly to the CPU and memory required to ingest data or perform transformation on the data. One OCU comprises 15 GB of RAM and 2vCPU. You will see one entry for compute in OCU-hours with the label for data ingestion. OCUs are billed on an hourly basis with per-minute granularity. You can configure a minimum and maximum number of OCUs per pipeline to control cost. Furthermore, OpenSearch Ingestion allows you to completely pause a pipeline when not in use and no OCUs are consumed when a pipeline is paused.

Amazon OpenSearch Service Vector Ingestion provides a simplified way to create Amazon OpenSearch Ingestion pipelines specifically for building vector databases on Amazon OpenSearch Service. You can optionally run vector auto-optimize jobs and use GPU-accelerated vector indexing to quickly optimize and build large-scale vector databases. There are no additional charges beyond the features you opt to use. Refer to the pricing for Amazon OpenSearch Ingestion, Amazon OpenSearch Service Vector Auto-Optimize Jobs, and Amazon OpenSearch Service Vector Index GPU-Acceleration for pricing respectively. Learn More

With Amazon OpenSearch Service Vector Index GPU-Acceleration, you only pay for resources consumed during accelerated vector indexing workloads. The compute capacity is measured in OpenSearch Compute Units (OCUs)—Vector Acceleration. This OCU type comprises 6GB of GPU memory, 8 GiB of RAM, and 2 vCPUs. When you have GPU-acceleration enabled, and your indexing throughput exceeds a configurable threshold, vector indexing tasks are offloaded to these OCUs. The OCUs will scale with higher throughput, but you pay based on total OCU time. In other words, if your OCUs are doubled, the processing time is halved, so your total cost is the same as without the doubling of OCUs. Auto-scaling and warm pooling is fully managed but isn’t added to your billed time. OCUs are billed hourly at the second granularity. Learn More about this feature.

With Amazon OpenSearch Service Vector Auto-Optimize Jobs, you pay a fixed rate for each job. Each job runs a hyperparameter optimization process. This involves sampling vectors from your dataset, launching parallel experiments to build and evaluate indexes using a variety of configurations, and generating a report with optimization recommendations. Learn More

Amazon S3 Vectors provides cost-effective, elastic, and durable vector storage at up to 90% lower costs for uploading, storing, and querying vectors. With S3 Vectors, you can power RAG and other semantic search workloads at scale, at a fraction of the cost for storage, requests, data uploaded, and data queried.

PUT cost
PUT pricing is based on logical GB of the vectors you upload, where each vector is the sum of its logical vector data, metadata, and key. You can upload multiple vectors in a single PUT request, maximizing upload throughput and minimizing upload costs.

Storage cost
Total storage is the sum of logical storage across your indexes, where the size of your storage is determined by the number of vectors you store and their size.  Vector size is determined by:

1) Vector data: Each vector has a size determined by number of dimensions. Each dimension equals 4 bytes of storage per vector, so for example, a 1024-dimensional vector requires 4 KB of logical vector data.

2) Metadata: You can store both filterable and non-filterable metadata with your vector. Non-filterable metadata is used to return information as a part of query results while filterable metadata can also be used to filter query results.

3) Key: Each vector is associated with a key. Keys require 1 byte of storage per character.

Query cost
Query charges include a per API charge in addition to a $/TB charge based on the average vector size, including vector data, key, and filterable metadata, multiplied by the number of vectors in the index you’re querying. As your vector index grows, data processing charges for query increase proportionally; however, at larger scale, you benefit from lower $/TB pricing above 100K  in your vector index.

With Amazon OpenSearch Service Direct Query, you only pay for the resources consumed by your workload. OpenSearch Service Direct Query charges for only the compute needed to query your connected data source as well as if you decide to index data in OpenSearch Service. The compute capacity is measured in OpenSearch Compute Units (OCUs). The number of OCUs corresponds directly to the vCPU and memory required to query or maintain indexes based on the data. One OCU comprises 2vCPU and 8 GiB of RAM. You will see one entry for compute in OCU-hours with the label for direct query. OCUs are billed on an hourly basis with per-minute granularity. To limit costs, you can set a maximum number of OCU-hours that can be used within a billing period using AWS Budget. If no queries or indexing jobs are active, no OCUs are consumed.

Costs for Direct Query OCUs will be based on the data size and frequency with which the indexed data in OpenSearch is kept updated. Serverless indexing, search, and managed storage costs will vary based on the size of the data indexed for use in the dashboards and the retention period in OpenSearch.

Now with direct query you can analyze logs data stored in Amazon S3, Amazon CloudWatch Logs, and Amazon Security Lake.

CloudWatch Logs customers building OpenSearch dashboards for VPC, WAF, CloudTrail

CloudWatch Logs customers can build OpenSearch dashboards in CloudWatch on their VPC, WAF and CloudTrail logs by navigating to "Analyze with OpenSearch" in CloudWatch Logs Insights, selecting the dashboard type and the logs. Prior to this step, CloudWatch customers first need to configure an OpenSearch integration - this step creates an OpenSearch collection which is used for storing the metrics needed for the logs in OpenSearch. Direct Query is used to keep the metric data in OpenSearch updated, by querying the CloudWatch logs and updating the metrics in OpenSearch, For instructions on how to create these dashboards refer to the CloudWatch documentation.

Amazon OpenSearch Service allows you to choose the type of Amazon EBS volume. If you choose Provisioned IOPS (SSD) storage, you will be charged for the storage as well as the throughput you provision. However, you will not be charged for the I/Os you consume.

UltraWarm is an Amazon OpenSearch Service tier allowing you to economically retain large amounts of data while keeping the same interactive analysis experience. Learn more »

Cold storage is the lowest-cost storage tier for Amazon OpenSearch Service, which lets you detach and store infrequently accessed data in Amazon S3 and pay for compute only when you need it. Learn more »

Note: Managed storage pricing is applicable to UltraWarm data, cold storage data and OR1 remote store data.

Amazon OpenSearch Service provides critical security fixes and operating system patches for engine versions that are in Extended Support, beyond the end of Standard Support, for a period of at least 12 months. This gives you more time to plan your upgrade to a more recent supported engine version. When you are running a version in Extended Support, you will be charged a flat fee/Normalized instance hour (NIH), in addition to the standard instance cost. NIH is computed as a factor of the instance size (e.g. medium, large), and number of instance hours. Please see the documentation for more information on Extended Support, calculating Extended Support charges, and the schedule for various versions. Please see below for Extended Support pricing per NIH.