OceanBase Database V4.3.3 is the first General Availability (GA) version in the V4.3 series. It achieves several key breakthroughs. First, it supports vector types and indexes applicable to AI-related analysis and processing, building on its foundation as a relational database. Second, it supports columnstore replicas deployed in a new form to meet requirements for strong isolation between transaction processing (TP) and analytical processing (AP) resources in hybrid transaction/analytical processing (HTAP) scenarios. Third, it significantly improves the performance of AP query tasks. OceanBase Database V4.3.3 also supports the complex ARRAY data type, improves the computing performance of the RoaringBitmap data type, enhances materialized view refresh, extends external table features, improves the import performance of external tables, and optimizes plan generation and execution strategies for AP SQL statements. These improvements enhance the capabilities for handling online analytical processing (OLAP) loads. Most features of OceanBase Database V4.2.4 and earlier are supported in V4.3.3. A new integrated version that applies to online transaction processing (OLTP) business scenarios will be released later.
This topic describes the key features and performance improvements of OceanBase Database V4.3.3.
Key features
Vector types and indexes (experimental)
The popularization of AI applications has triggered explosive growth in unstructured data such as images, videos, and texts. In this case, vector databases emerged accordingly. A vector database provides a fully managed solution for storing, indexing, and retrieving high-dimensional vectors. It can convert a keyword-based search into a vectorized retrieval to support large-scale similarity searches for high-dimensional vectors.
OceanBase Database V4.3.3 in MySQL mode supports vector type storage, vector indexes, and embedding vector retrieval. It supports float vectors with at most 16,000 dimensions, basic operations such as addition, subtraction, multiplication, comparison, and aggregation, as well as precise search and Approximate Nearest Neighbor Search (ANNS), along with Hierarchical Navigable Small World Network (HNSW) indexes for at most 2,000 dimensions. It adapts to business scenarios such as image and video retrieval, behavior recommendation, as well as security and fraud detection.
For more information, see Overview.
ARRAY is a common complex data type in AP business scenarios. An array can store multiple elements of the same type. You can use it to represent multi-valued attributes. OceanBase Database V4.3.3 supports the ARRAY data type in MySQL mode. During table creation, you can define a column as an array of numeric or character values, which can also be an embedded array. You can also create an expression for querying or writing arrays. You can use the array_contains expression and the ANY operator to verify whether an array contains a specific element. You can also use operators, such as +, -, =, and !=, to calculate the elements in an array.
For more information, see Overview.
Read-only columnstore replicas (experimental)
OceanBase Database V4.3.0 and later support columnar storage. To implement physical isolation between TP resources and AP resources in HTAP scenarios, OceanBase Database V4.3.3 introduces a new deployment form that allows you to add a separate zone for storing read-only columnstore replicas based on the original cluster. In the new zone, all user tables are stored in columnar storage mode. In an AP business scenario, you can use an independent OceanBase Database Proxy (ODP) for database connection and set the session-level system variable ob_route_policy, which specifies the routing strategy, to COLUMN_STORE_ONLY so that you can access only columnstore replicas for queries and analysis in weak-consistency read mode, without affecting TP business.
For more information, see Columnstore replica.
This feature is in the experimental stage in OceanBase Database V4.3.3, and will be enhanced for use in production environments in later versions.
Enhancements in materialized view capabilities
OceanBase Database V4.3.1 and later support materialized view-based rewriting. When the system variable QUERY_REWRITE_ENABLED is set to True, you can specify the ENABLE QUERY REWRITE clause in the materialized view creation statement to enable automatic rewriting, to reduce the manual rewrite workload. OceanBase Database V4.3.3 also supports query rewriting in the following scenarios when specific conditions are met: (1) The join order or join type in the FROM clause of the original query is different from that of the materialized view. (2) The original query contains tables not included in the materialized view. (3) The original query needs to be rewritten into a query based on an aggregate materialized view. (4) Aggregation rollup is involved.
For more information, see Rewrite queries based on materialized views (MySQL mode) and Rewrite queries based on materialized views (Oracle mode).
OceanBase Database supports only rowstore materialized views in earlier versions. OceanBase Database V4.3.3 also supports columnstore materialized views, making it possible to achieve higher query performance in complex analysis scenarios that involve materialized view references. For more information, see Create a materialized view (MySQL mode) and Create a materialized view (Oracle mode).
Distributed late materialization
The late materialization strategy optimizes the order of index table-based filtering and data queries from the primary table, thus reducing performance resource consumption by random I/O operations. OceanBase Database V4.3.3 supports late materialization for partitioned tables.
Enhanced INSERT OVERWRITE statement
OceanBase Database V4.3.2 and later support table-level overwriting based on the INSERT OVERWRITE statement. In OceanBase Database V4.3.3, you can specify partitions, subpartitions, or columns of the destination table in the INSERT OVERWRITE statement, allowing for more flexibility in data overwriting and making it applicable to a wider range of business scenarios.
For more information, see Insert data (MySQL mode) and Insert data (Oracle mode).
High performance
Performance optimization for the RoaringBitmap data type
OceanBase Database V4.3.3 focuses on analyzing the performance issues of RoaringBitmap data computations and significantly improves the execution performance in cardinality, AND, OR, XOR, ANDNOT, and aggregate scenarios by optimizing memory application and expression execution logic to reduce the unnecessary performance overhead.
High reliability
Quick restore
OceanBase Database V4.3.3 provides a quick restore feature that allows you to restore only logs rather than data. The data backup feature allows you to build an intermediate-layer index for a backup SSTable based on the backup address. With this index, OBServer nodes can randomly read data from the backup SSTable like reading local data. This way, you can use data read and write services without restoring data to your local server.
Ease of use
Log stream replica management
OceanBase Database V4.x replaces the concept of partition with log stream. OceanBase Database V4.3.3 redesigns the partition O&M commands in earlier versions for log stream replica-level O&M. It provides a series of syntaxes for adding log stream replicas, dropping log stream replicas, converting the type of log stream replicas, migrating log stream replicas, modifying the number of Paxos members of a log stream replica, and canceling a disaster recovery task, to meet requirements for manual log stream replica O&M.
References
This topic describes new and key features of OceanBase Database V4.3.3. For more information about other versions, see the following release notes: