GBase 8s Database: The Four Weapons — Part 3: The "Golden Shield" — High Availability Architecture

In mission-critical sectors such as finance, telecommunications, and energy, where systems must run 7×24 without interruption, database availability must reach the stringent 99.999% (five nines) standard. GBase 8s delivers a comprehensive high-availability technology stack — four layers of protection that scale from single data center to cross-region deployments, and from sub-second recovery to near-transparent failover. This provides core business systems with a multi-tiered, customizable "never down" assurance framework. In this installment of our "Four Weapons of 8s" series, we break down the protective power of this "Golden Shield."

Layer 1: Primary-Standby Cluster
Local and Same-City "Shadow Guards"

GBase 8s offers a primary-standby cluster based on real-time log synchronization. The core principle is straightforward: every operation performed on the primary node generates an operation log (redo log), which is synchronously transmitted to the standby node. The standby node replays these logs to maintain data consistency. Two modes address different scenarios:

HAC Cluster (Same-City Primary-Standby): One primary, one standby — directly comparable to Oracle Active Data Guard. Three synchronization modes allow flexible trade-offs between data consistency and performance:

Fully Synchronous: The primary node's transaction commits only after the standby node confirms redo completion. Slower, but guarantees zero data loss (RPO = 0). Ideal for accounting systems and other scenarios with extreme consistency requirements.

Near-Synchronous: The primary commits as soon as the standby acknowledges log receipt. Balances speed and safety — suitable for most core transaction workloads.

Asynchronous: The primary commits immediately and syncs with the standby in the background. Delivers maximum performance for query/analytics workloads where minor data latency is acceptable.

Deployment is straightforward, failover typically completes within 10 seconds, and both primary and standby nodes support read/write operations, effectively distributing query load.

RHAC Cluster (Remote Disaster Recovery): An asynchronous one-primary-multiple-standby cluster designed specifically for cross-city disaster recovery. Given the inherent latency of long-distance networks, RHAC does not require real-time acknowledgment, minimizing impact on primary node performance and comfortably handling the latency and bandwidth constraints of remote transmission.

Layer 2: Shared Storage Cluster (SSC)
The "Multi-Node Powerhouse" for High-Performance Core Systems

For mission-critical workloads with massive data volumes (typically 10 TB to 100 TB) and demanding performance and scalability requirements, GBase 8s offers a different approach: multiple nodes share a single storage instance. All nodes support read/write operations, queries access local cache with zero network overhead, and data is stored only once — reducing storage costs while ensuring data consistency.

This architecture is directly comparable to Oracle RAC, with excellent linear scalability and peer-to-peer management. When a single node fails, other nodes take over within 5 seconds — virtually transparent to business operations. For systems under peak transaction pressure, multiple nodes serving requests simultaneously deliver both higher performance and greater reliability.

Layer 3: Enterprise Replication (ER)
The "Site-Agnostic" Flexible Synchronization Solution

While the first two layers address server- or data center-level failover, ER handles flexible data flow at the logical level. Unlike cluster solutions that require shared storage or full data synchronization across all nodes, ER replicates data at the table level — comparable to Oracle GoldenGate:

Supports one-way or bidirectional synchronization across multiple nodes, enabling active-active architectures where two nodes act as mutual primaries and synchronize data with each other.

Can replicate full tables or specific columns only, with support for full initialization, incremental synchronization, and resumable transfer.

Failover completes within 1 second.

The flexibility of this solution lies in its minimal infrastructure requirements — it does not require high-speed dedicated links or shared storage between nodes. As long as network connectivity exists, ER works. This makes it ideal for data aggregation between provincial and municipal/county levels, data sharing across parallel organizations, or remote unattended base stations requiring minimal human intervention.

Layer 4: The Combined Approach
The Ultimate "Two-City, Three-Center" Solution

For the most demanding core systems with extreme availability requirements, no single layer of protection is sufficient. GBase 8s supports combining the above technologies into a mature "Two-City, Three-Center" architecture.

A typical deployment: a 4-node SSC cluster in the primary data center handles high-performance read/write operations; an HAC node in the same-city disaster recovery center provides real-time data protection via near-synchronous or fully synchronous mode; and an RHAC node in a remote disaster recovery center guards against regional catastrophes such as earthquakes. All node access is unified through the Connection Manager (CM), shielding the application layer from backend architectural details. Any server failure or data center outage triggers automatic failover within the predefined time window.

From flexible primary-standby clusters, to high-performance shared storage clusters, to fine-grained data synchronization and multi-center disaster recovery, GBase 8s delivers a clearly structured, flexibly composable high-availability product stack. For enterprises undertaking migration and modernization initiatives, this means access to functionally equivalent — and in some scenarios, superior — technology alternatives. Domestic databases now have the proven capability to support the "never-stop" operation of critical national infrastructure systems.