Abstract: This paper introduces the key features and successful cases of GaussDB(openGauss).
GaussDB(openGauss) is a new-generation enterprise-level distributed database that integrates Huawei’s experience in the database field and meets enterprise-level requirements. It supports both centralized and distributed deployment modes.
It is designed for financial Internet transactions and government and enterprise OA/ office scenarios, with the advantages of security, reliability, ultra-high performance and ease of use.
Shen Zhipeng, senior product manager of Huawei Cloud Database Product Department, introduced the key features and success cases of GaussDB(openGauss) in depth on the live broadcast of Programmer 1024.
Distributed or centralized, one is for you
GaussDB(openGauss) is positioned as an enterprise-level cloud distributed database. In terms of architecture, it focuses on the enterprise-level capabilities of traditional databases and the high scalability and availability of Internet distributed databases.
Compared with the traditional mode, GaussDB(openGauss) can obtain:
High performance: strong consistency of distributed transactions is supported, and the measured performance index of 12 million TPMCS can be achieved under 32 nodes;
High availability: Supports multi-active DISASTER recovery (Dr) across azs in the same city and financial deployment in two places and three centers.
High expansion: currently supports online expansion of up to 256 nodes, while ensuring customers have excellent linear performance growth, breaking the performance ceiling of traditional DIY architecture;
Easy operation and maintenance: highly support HTAP mixed load scenarios, greatly reduce the cost of business transformation, and obtain online monitoring, operation and maintenance, upgrade and other convenient capabilities based on the cloud platform.
GaussDB(openGauss) can be deployed in two modes on huawei cloud: centralized and distributed. These modes are applicable to enterprise core transactions and future massive transaction scenarios respectively to enhance differentiated competitiveness.
Centralized deployment
Centralized deployment includes single-node deployment and active/standby deployment.
For example, active/standby mode supports 1+2 (maximum protection) active/standby mode. Hot backup based on database log replication provides high availability when the performance of a single machine meets requirements.
1+1 (maximum available) indicates that data is written to the standby server at the same time. However, if the network is affected, the synchronization operation cannot be completed and the system changes to asynchronous mode. If the network is recovered, the system will automatically catch up with the system. During data synchronization, data will be lost during the switchover.
1+2 (Maximum protection) indicates that data is written to the standby machine at the same time and must be acknowledged before being returned to the client. High reliability.
The centralized version has an open source ecosystem, and users can download it directly from the open source website. As the only open source database in China, it is also the best proof that Huawei is an open source, open and non-lockin single vendor.
Distributed deployment:
In distributed deployment, data is divided by Shard, and the read and write loads expand quasi-linearly, meeting the requirements of large-scale service scenarios and supporting the high availability deployment in the two centers and three centers. In addition, the distributed version carries huawei cloud self-developed distributed component system, which is a strong guarantee for traditional enterprises to embrace the Internet and face the challenges of future massive transactional scenarios.
Here are some key players in distributed architecture:
Distributed architecture can be divided into independent deployment and hybrid deployment.
The mixed deployment scheme is suitable for general customers. The scheme includes: 1. Three copies of each role and three copies of data are deployed; 2. All roles and processes are deployed in one system, and only the database node is displayed externally.
The advantages of this deployment are simple networking and efficient delivery interface. The starting point configuration requirements are low, and the adaptation scenarios are common. It matches the future evolution of technology.
Another independent deployment scheme suitable for high-end customers includes: 1. Three copies of each role and three copies of data are deployed; 2. Key role processes are deployed separately. CMS, GTM, CN, and DN are active and standby.
In independent deployment mode, users can determine the optimal ratio of CN and DN based on service loads to achieve the most efficient networking.
Explains the key distributed features of GaussDB(openGauss)
Key technology 1: distributed execution framework
A service application delivers SQL to a Coordinator. SQL can be used to add (INSERT), delete/drop, update, and select data. A Coordinator uses the optimizer of a database to generate an execution plan. Each DN processes data according to the execution plan.
Since data is evenly distributed on each node using the consistent Hash technology, the DN may need to obtain data from other DN during data processing. GaussDB provides three stream streams (broadcast stream, aggregation stream, and redistribution stream) to reduce data flow among DN nodes.
DN returns the result set to Coordinate for summary. The Coordinator returns the summary result to the service application.
Key technology two: GTM-Lite technology
Gtm-lite technology can ensure global transaction consistency while providing high performance transaction processing capability, avoiding the performance bottleneck of single GTM.
High performance transaction management here refers to the technique of no-lock, multi-version, high-concurrency transactions.
Moreover, the distributed GTM-Lite solution provides global transaction snapshot and commit number management, achieving strong consistency without central node performance bottlenecks.
Key technology three: High performance transaction processing based on NUMA-Aware
Numa-aware database architecture based on Kunpeng CPU has better performance.
Key technology four: Cluster HA, multi-layer redundancy to achieve system without single point of failure
GaussDB(openGauss) implements hardware redundancy, instance redundancy, and data redundancy to ensure high availability without single points of failure.
High availability of hardware includes RAID redundancy for storage disks, redundancy for dual switches, redundancy for multiple network adapters, and UPS power protection for hosts.
High software availability includes CN instances of coordination nodes in active-active redundancy, data nodes/global transaction managers/cluster manager instances in active-standby redundancy.
In addition, it also supports software and hardware failure detection to further ensure the high availability of the database.
Key technologies 5: Cross-AZ /Region Dr Brings high availability
HA in the cluster ensures that data is not lost and services are interrupted in seconds. Same-city, cross-AZ Dr, data loss, minute-level recovery, and geo-redundant deployment.
In addition to cross-az in the same city and geo-centric, GaussDB(openGauss) is also exploring remote multi-living, which has the following features:
- Multi-center unified query and global consistent read, high overall resource utilization;
- Flexible HA solution: Multiple copies can be configured to implement DC, AZ, and Region-level HA Dr Policies.
- Load sharing and seamless fault switching;
- Supports smooth online capacity expansion.
Key technology six: Scale-out online horizontal expansion brings high expansion
GaussDB(openGauss) supports a maximum of 256 nodes in a cluster. When nodes are expanded, excellent linear performance ratio can be achieved
At the same time, node expansion can be completely online and transparent to customers. After node expansion, DN data is automatically redistributed and CN load balancing is automatically completed.
Finally, in addition to bringing GaussDB(openGauss) distributed key feature interpretation, during the 1024 Programmer Carnival, Huawei cloud also prepared a large wave of technology dry goods live activities, can not only learn technology, but also blind box, koi gift package.
On October 23, we invited huawei cloud native open source head, Huawei Cloud DevCloud chief technology evangelist and other 10+ big players to appear, analyze the industry trend of cloud native, inclined to teach cloud native practice secrets. Click to watch live.
There are dry goods through train, take you to reveal the deepest code technology of big factory, click to view each technical venue, dig more dry goods.
Click to follow, the first time to learn about Huawei cloud fresh technology ~