AnalyticDB for MySQL uses a cloud-native architecture and separates computing resources from storage resources as well as hot data from cold data. It supports real-time data write operations with high throughput, strong data consistency, and hybrid load of highly concurrent queries and batch processing with high throughput.
AnalyticDB for MySQL separates the access layer, computing layer, and storage layer. Each layer can be scaled separately. All service nodes are in the serverless state, and all nodes are in Active-Active or Primary-Standby mode. When software or hardware failures occur, the scheduling system can automatically detect and migrate data of faulty nodes. This process runs without interrupting your applications. AnalyticDB for MySQL features a fully distributed architecture that covers request routes, storage, and computing. This can prevent the system from becoming unavailable due to a single faulty node.
AnalyticDB for MySQL supports online scale-out and rolling upgrade. The storage space and calculation amount can be increased from GB level to PB level. AnalyticDB for MySQL supports automatic data distribution and automatic index construction without the need to shut down servers. This allows you to purchase services and scale clusters as your business grows.
AnalyticDB for MySQL supports two levels of partitions. Level-1 partitions are divided into shards. Level-2 partitions are specified within shards. You can use distribution keys to specify multiple shards and use partition keys to specify level-2 partitions from a specific dimension. AnalyticDB for MySQL allows you to store data in multiple replicas. Each replica supports read and write operations. Read-only replicas will be available soon.
AnalyticDB for MySQL allows you to create an index for multiple columns and multiple conditions. This can help you quickly find result sets that meet the conditions. Except for traditional indexes such as inverted indexes, bitmaps, and range trees, AnalyticDB for MySQL also supports full-text indexes, JSON indexes, and vector indexes. As incremental data is written, indexes can be automatically restructured and merged. This process does not require shutdown of servers or cause interruptions into your applications.
AnalyticDB for MySQL supports the rule-based optimizer (RBO) and provides hints to modify and intervene in execution plans. Hints can be placed anywhere in an SQL statement and can affect query plans of the specified blocks to be queried based on different positions. AnalyticDB for MySQL supports execution plan optimization based on costs. Optimizers can attempt various execution plans to enable global optimization. AnalyticDB for MySQL can generate optimal execution plans for data of different sizes based on data characteristics.
AnalyticDB for MySQL supports new technologies such as vectorized execution and code generation as well as accelerated hardware such as graphics processing units (GPUs) and field programmable gate arrays (FPGAs).
The vector analysis feature of AnalyticDB for MySQL allows you to query and analyze similarities in unstructured data. AnalyticDB for MySQL uses an AI algorithm to extract the features of unstructured data, and then uses a feature vector to uniquely identify the unstructured data. The distance between vectors is used to measure the similarities between unstructured data. The vector analysis feature uses a full-index structure that allows you to use SIMD instruction acceleration, efficient indexing algorithms, hybrid retrieval cost-based optimizer (CBO) policies, and low-cost storage technologies to enable approximation queries and analysis of unstructured data at high performance and low costs. The vector analysis feature allows you to directly read unstructured data such as images, audio files, and videos in storage devices such as Object Storage Service (OSS). Then, AnalyticDB for MySQL generates vectors for the data, provides vector indexes, and uses SQL statements to analyze and query unstructured data.
The Cluster Edition scales nodes and storage space within seconds. Data query is not affected when you change the cluster specifications.
Edition Model Description Feature Performance vCPUs Memory (GB) Disk space (GB) Cluster Edition (compute-intensive) C8 8 64 100 to 1,000 Allows you to scale out nodes and scale up storage space. Provides 100 times the performance of MySQL databases and offers linear scalability. C24 24 96 100 to 1,000 C4 4 32 100 to 200 Allows you to scale out nodes. Cluster Edition (storage-intensive) S8 8 64 1,000 to 12,000 Allows you to scale out nodes and scale up storage space. Provides slightly lower performance compared with compute-intensive node groups and offers linear scalability.
- Large-scale database clusters featuring high read and write speeds
The Cluster Edition synchronizes replicas in compliance with the Raft consensus protocol and creates indexes in a lightweight manner. This achieves high throughput for read and write operations in real time. The Cluster Edition improves the distributed hybrid computing engine and the query optimizer to provide a stronger complex computing capability.
- Higher availability and reliability
The Cluster Edition allows you to deploy clusters within a zone or across zones. AnalyticDB for MySQL features high availability (at least 99.95%) and can resume services within seconds upon failure. It can also automatically detect faults, remove faulty nodes, and rebuild replicas. It stores data in three replicas and backs up full and incremental data on a regular basis. It can provide you with the data reliability required in the finance industry.
- Sophisticated analysis of MySQL data
The Cluster Edition supports sophisticated analysis of MySQL data and real-time integration with business data to provide you with data-driven insights into your business.
- Real-time data warehouses for large and medium-sized enterprises
The Cluster Edition uses databases to build enterprise-level real-time data warehouses in a simple and cost-effective manner. AnalyticDB for MySQL supports multiple data sources and enhanced data extraction, transformation, and loading (ETL). Compared with offline data warehouses, this solution reduces costs by 40%.
- Backflow acceleration and interactive analysis of big data in Hadoop, EMR, and Spark
The Cluster Edition supports backflow acceleration of data in Hadoop, EMR, and Spark to facilitate data migration. It supports flexible and quick SQL query and multidimensional analysis of trillions of wide tables within milliseconds.