Managing capacity limits for Amazon OpenSearch Serverless
With Amazon OpenSearch Serverless, you don't have to manage capacity yourself. OpenSearch Serverless automatically scales the compute capacity for your account based on the current workload. Serverless compute capacity is measured in OpenSearch Compute Units (OCUs). Each OCU is a combination of 6 GiB of memory and corresponding virtual CPU (vCPU), as well as data transfer to Amazon S3. For more information about the decoupled architecture in OpenSearch Serverless, see How it works.
When you create your first collection, OpenSearch Serverless instantiates a total of four OCUs (two for indexing and two for search). These OCUs always exist, even when there's no indexing or search activity. All subsequent collections can share these OCUs (except for collections with unique AWS KMS keys, which instantiate their own set of four OCUs). If needed, OpenSearch Serverless automatically scales out and adds additional OCUs as your indexing and search usage grows. When traffic on your collection endpoint decreases, capacity scales back down to the minimum number of OCUs required for your data size. For the search and time series collection, the number of OCUs required when idle is proportional to the data size and index count. For vectors, it depends on both the memory (RAM) to store vector graphs and disk space to store indices. If not in an idle state, OCU requirements take both of these into account.
Vector collections keep index data in OCU local storage. OCU RAM limits are reached faster than OCU disk limits, causing vector collections to be restricted by RAM space. At most, it will scale down to 1 OCU [0.5 OCU x 2] for indexing and 1 OCU [0.5 OCU x 2] for search. Scaling also factors in the number of shards needed for your collection or index. Each OCU can support a specified number of shards. The number of indexes should be proportional to the shard count. The total number of base OCUs required is the maximum amount of your data, memory, and shards required. To learn more, see Amazon OpenSearch Serverless cost-effective search capabilities, at any scale.
For search and vector search collections, all data is stored on hot indexes to ensure fast query response times. Time series collections use a combination of hot and warm storage, keeping the most recent data in hot storage to optimize query response times for more frequently accessed data. For more information, see Choosing a collection type.
Note
A vector search collection can't 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 collection. The OCUs of vector collections are shared among the same KMS key collections.
To manage capacity for your collections and to control costs, you can specify the overall maximum indexing and search capacity for the current account and Region, and OpenSearch Serverless scales out your collection resources automatically based on these specifications.
Because indexing and search capacity scale separately, you specify account-level limits for each:
-
Maximum indexing capacity – OpenSearch Serverless can increase indexing capacity up to this number of OCUs.
-
Maximum search capacity – OpenSearch Serverless can increase search capacity up to this number of OCUs.
Note
At this time, capacity settings only apply at the account level. You can't configure per-collection capacity limits.
Your goal should be to ensure that the maximum capacity is high enough to handle spikes in workload. Based on your settings, OpenSearch Serverless automatically scales out the number of OCUs for your collections to process the indexing and search workload.
Configuring capacity settings
To configure capacity settings in the OpenSearch Serverless console, expand Serverless in the left navigation pane and select Dashboard. Specify the maximum indexing and search capacity under Capacity management:
To configure capacity using the AWS CLI, send an UpdateAccountSettings request:
aws opensearchserverless update-account-settings \ --capacity-limits '{ "maxIndexingCapacityInOCU":8,"maxSearchCapacityInOCU":9}'
Maximum capacity limits
The maximum total of indexes a collection can contain is 1000. For all three types of collections, the default maximum OCU capacity is 10 OCUs for indexing and 10 OCUs for search. The minimum OCU capacity allowed for an account is 1 OCU [0.5 OCU x 2] for indexing and 1 OCU [0.5 OCU x 2] for search. For all collections, the maximum allowed capacity is 500 OCUs for indexing and 500 OCUs for search. You can configure the OCU count to be any number from 1 to the maximum allowed capacity, in multiples of 2.
Each OCU includes enough hot ephemeral storage for 120 GiB of index data. OpenSearch Serverless supports up to 1 TiB of data per index in search and vector search collections, and 30 TiB of hot data per index in a time series collection. For time series collections, you can still ingest more data, which can be stored as warm data in S3.
For a list of all quotas, see OpenSearch Serverless quotas.
Monitoring capacity usage
You can monitor the SearchOCU and IndexingOCU account-level
CloudWatch metrics to understand how your collections are scaling. We recommend that you
configure alarms to notify you if your account is approaching a threshold for metrics
related to capacity, so you can adjust your capacity settings accordingly.
You can also use these metrics to determine if your maximum capacity settings are appropriate, or if you need to adjust them. Analyze these metrics to focus your efforts for optimizing the efficiency of your collections. For more information about the metrics that OpenSearch Serverless sends to CloudWatch, see Monitoring Amazon OpenSearch Serverless.
