Data consistency
Due to the read/write splitting architecture of DashVector, a document may not be found in a search or obtained immediately after being inserted, updated, or upserted; similarly, a document may still be found in a search and obtained immediately after being deleted.
DashVector ensures final data consistency. That is, all operations on documents are performed in sequence.
Limits on free trial clusters
If you need to adjust the limits on free trial clusters, such as those on the collection quantity, document quantity, and QPS, or if you have any requirements or feedback, or want to obtain more technical support, feel free to contact us in the following ways:
Official DingTalk group: 25130022704
Email address for technical support: dashvector@service.aliyun.com
Item | Upper limit |
Number of free trial clusters that can be created under each Alibaba Cloud account | 1 |
Number of collections that can be created in a free trial cluster | 2 |
Validity period of a free trial cluster | One month. You can create another free trial cluster after the existing one expires or is released. |
QPS for obtaining documents |
|
QPS for deleting documents |
|
QPS for inserting documents |
|
QPS for updating documents |
|
QPS for upserting documents |
|
A collection in a free trial cluster has a fixed storage capacity. The maximum number of documents allowed in a collection varies according to the number of vector dimensions. In addition, the QPS for searching for documents varies according to the number of vector dimensions and is 30 for a free trial cluster by default. You can estimate the maximum number of documents allowed and the QPS for searching for documents by referring to the following table.
Number of vector dimensions | Theoretical maximum number of documents allowed in a collection per free trial cluster | Theoretical QPS for searching for documents in a collection per free trial cluster |
64 | 2,400,000 | 30 |
128 | 1,200,000 | 30 |
256 | 600,000 | 30 |
512 | 300,000 | 15 |
768 | 200,000 | 10 |
1,024 | 150,000 | 7 |
1,536 | 100,000 | 5 |
2,048 | 75,000 | 3 |
4,096 | 37,500 | 1 |
8,192 | 18,750 | 1 |
16,384 | 9,375 | 1 |
20,000 | 7,680 | 1 |
Limits on a paid cluster
The capacity and QPS of a paid cluster are only subject to factors such as the cluster specifications, data volume, and number of vector dimensions. For more information, see Specifications.
Paid clusters support a maximum of 32 Collections. The number of Partitions that can be created scales with the cluster tier and is independent of the number of Collections.
Taking P.2xlarge as an example, a cluster of this tier supports 2048 Partitions. If only one Collection exists, the Collection can create up to 2048 Partitions. With 32 Collections, each Collection can create a maximum of 64 Partitions on average.
Cluster type | Cluster specification | Number of collections that can be created | Number of partitions that can be created |
Performance-optimized type | P.small | 32 | 512 |
P.large | 32 | 1024 | |
P.2xlarge | 32 | 2048 | |
P.4xlarge | 32 | 4096 | |
P.8xlarge | 32 | 8192 | |
P.16xlarge | 32 | 16384 | |
Storage-optimized type | S.small | 32 | 512 |
S.large | 32 | 1024 | |
S.2xlarge | 32 | 2048 | |
S.4xlarge | 32 | 4096 | |
S.8xlarge | 32 | 8192 | |
S.16xlarge | 32 | 16384 |
General limits
Item | Upper limit |
Number of API keys that can be created under each Alibaba Cloud account | 3 |
Collection name | It must be 3-32 characters in length, and can contain letters, digits, underscores (_), and hyphens (-) . |
Number of vector dimensions | (1, 20000] |
Partition name | It must be 3-32 characters in length, and can contain letters, digits, and underscores (_) . |
Field name | It must be 1-32 characters in length, and can contain letters, digits, and underscores (_) . |
Document ID | It must be 1-64 characters in length, and can contain letters, digits, and the following special characters: _ - ! @ # $ % + = . |
Number of documents that can be obtained at a time |
|
Number of documents that can be deleted at a time |
|
Number of documents that can be inserted at a time |
|
Number of documents that can be updated at a time |
|
Number of documents that can be upserted at a time |
|
Top k documents that can be returned in a document search | 1,024 |
Grouped vector search maximum group_count | 64 |
Grouped vector search maximum group_topk | 16 |
Request body size | 2 MB. Recommended: <= 50 KB. |
Number of fields | 1,024. Recommended: <= 64. |