GaussDB

Strong consistency between clusters: GaussDB uses decoupled storage and compute and is equipped with Kunpeng processors, NoF networking, and Dorado storage to provide full-stack high availability. Cluster-level faults are completely isolated in the dual-AZ active-active system, achieving an RPO of zero.

Lossless switchover transparent to applications: Databases are instantly connected after a failover, and SQL operations are resumed from the breakpoint, where workloads were not affected. This ensures database service continuity and enables automatic transaction replay during and after HA switchovers/failovers.

Data reliability: Twelve nines (99.9999999999%) of data durability guarantee data security and reliability.

Full software encryption: GaussDB supports SM series cryptographic algorithms and transparent data encryption. Such algorithms allow data to be queried and calculated in the ciphertext. This enables GaussDB to provide more than 35% higher ciphertext data processing efficiency than similar products.

Anti-tampering: GaussDB uses a high-concurrency digest generation algorithm to generate table-level hash values. This algorithm provides support for 10x more concurrency than when serial processing is used. It supports SQL and complex operations of associating multiple tables or ledgers.

Application development: GaussDB provides full-link, full-data SQL awareness, analysis, and optimization to help you develop applications more easily and efficiently.

O&M: GaussDB ensures quick, precise intelligent O&M based on full-process monitoring and intelligent diagnostics.

Intelligent approaches, including index recommendations, distribution column recommendations, and root cause analysis, improve diagnosis efficiency by more than five times.

In-place updates: With the high-performance storage engine, GaussDB responds fast all day long with almost no jitter. The performance jitter remains lower than 3% even when handling heavy workloads.

Online cluster scale-out and multi-round incremental data synchronization: Multiple rounds of incremental data synchronization are performed to reduce table locking. System capacity and performance can be scaled out online in just a few seconds.

Flexible, lightweight deployment: You can create a DB instance consisting of only one primary node, one standby node, and one log node to conserve storage and compute resources. The deployment costs can be reduced by 1/3.

Refined management: Resources are assigned and isolated by tenant, and even an individual vCPU can be managed, improving the overall resource efficiency.

One-stop migration solution: GaussDB uses UGO's pre-migration evaluation and schema migration techniques to automatically convert syntax of mainstream databases to GaussDB syntax, and uses DRS to migrate heterogeneous database data online.

A traditional row storage engine works using append updates, which means if there are frequent updates, a lot of junk data is generated and data access efficiency is reduced. Additionally, deleting junk data consumes a large number of resources, causing performance jitter.

The storage engine used by GaussDB uses in-place updates, where valid data of the most recent version and junk data of historical versions are stored separately. This prevents the data space from expanding and increases garbage collection efficiency. The performance jitter is less than 3% even in the face of a massive volume of concurrent requests, and storage utilization is improved by 17%.

Two sets of independent database software are deployed in separate DCs as two logically isolated clusters. If a fault occurs in one DC, workloads can be quickly switched to the standby DC, which will not be affected by the fault. The service downtime is less than 2 minutes.

GaussDB uses storage replication to implement an intra-city dual-cluster DR solution. With the outstanding performance and reliability of shared storage devices, the RPO of cross-cluster DR is zero.

Data is processed in ciphertext during storage, transmission, and query.

Database users keep keys in their hands. Even database administrators do not have access. Encryption and decryption are performed only on the user side.

Syntax is automatically parsed without affecting applications. Syntax parsing is embedded in the driver, so the original SQL statements and data types do not need to be modified.

Some algorithms are not involved in encryption and decryption, which reduces hardware I/O.

Mathematical algorithms are used to enable direct query in ciphertext, minimizing performance loss during encryption and decryption.