TBase Tencent data platform team PostgreSQL open source on the basis of research and development of enterprise-class distributed database management system HTAP:
-
High performance scalable distributed transaction capabilities to support two kinds of RC and RR isolation level;
-
By security, management, audit separation of powers system, provide a full range of data security assurance mechanisms;
-
High-performance partition table, so that the data retrieval efficiency can increase exponentially;
-
SQL 2003 standard-compliant aspects, PostgreSQL syntax and common functions & Oracle data types, functions, and so the window;
-
Efficient data management capabilities to provide business data size separation, data separation cold
TBase architecture:
There are three types of nodes in the cluster, each assume different functions into a system through a network connection. The three node types are:
-
Coordinator: coordinator node, provide external interfaces for data distribution and query plans, such as the position of the plurality of nodes, each node provides the same view of the database, CN global metadata storage system.
-
Datanode: storing metadata associated processing node, each node also stores a data slice. Functionally, DN node is responsible for the completion of execution request coordinator node distribution.
-
GTM: Global Transaction Manager (. Global transaction manager), responsible for managing the cluster transaction information, while managing a cluster of global objects, such as sequence, does not provide other functions in addition to the GTM.
TBase Features:
-
Distributed global transaction consistency capability: a distributed transaction consistency by own patented technology, including two-phase commit (Two Phase Commit) and a global clock (Global Timestamp) strategies to ensure full transactional consistency in a distributed environment .
-
SQL compatibility: various types of NoSQL SQL2003 standard, PostgreSQL syntax, commonly used functions & Oracle data types, UDF / UDAF, common window function, JSON / JSONB / XML / arrays, etc., recursive WITH, no lock DDL operations, extensions and so on.
-
HTAP capacity: providing two planar view OLTP and OLAP, OLTP datanode traffic running on the master node, OLAP services running on the standby node datanode node, Secret improve website weight of both synchronous data stream by way replication is performed.
-
Separate read and write capabilities: Providing two planar view write and read-only, read-write request traffic processed by the master node, read-only request traffic processed by the standby node, the standby node synchronization data of the main flow by way replication is performed .
-
Superior data security capabilities: through the system of separation of powers, the role of the traditional database system DBA is broken down into three separate roles: security administrator, audit administrator, data administrator; based on this proposed security policy, mainly fine divided into three parts: data encryption, data access desensitization, mandatory access control, the three combined to provide multiple levels of data security capabilities.
-
Efficient data management capabilities: the inclination data management, storage and performance to solve the uneven distribution of pressure data caused; cold hierarchical storage, reduce storage costs for business, improve thermal performance data.
-
Multi-core parallel computing capabilities: internal nodes using parallel computing, according to the size of the table to start multiple processes to collaboratively a query.
-
Multi-tenancy capability: based on intra-node cluster group node group of multi-tenant solutions, services and resources to achieve isolation internal database cluster, multiple business within TBase isolated operation.
-
Multi-level disaster recovery capabilities: using a strong master synchronous replication to ensure data from exactly the same, when the primary node failure without loss of data protection; provide based recovery characteristics at any point in time to prevent data loss caused by misuse.
-
Online capacity expansion: by introducing a shard map layer (per one stored mapping relationship between the DN and shardid shard map) is, when a new node added, just need to put some shardid shard map mapping to the new node is added, and to the corresponding the data will be able to move past, greatly reducing the time expansion.
-
Surrounding rich ecological capacity: PostGIS, heterogeneous data replication, LVS load balancing, FDW federal capacity.
TBase application scenarios:
-
In the localization of landing, the core business areas of the scene to O;
-
There are business scenarios need to HTAP mixed model analysis of application transactions;
-
There are business scenarios demand for things Geographic Information System;
-
对实时高并发环境下的分布式事务强一致性有要求的业务（如金融、证券等业务）场景；
-
对企业级数据安全、审计、治理应用相关功能有需求的业务场景；
-
对高可用、异地容灾、7*24 小时服务能力有需求的业务场景；
-
有海量存储计算需求、弹性扩容、在线扩容需求的业务场景；
-
对异构数据互通共享有需求的业务场景。