Choosing a database deployment layout is not just an installation decision. It determines where infrastructure costs accumulate, how failures affect availability, and how easily the cluster can scale in the future.
GBase Database(GBase 8c) supports different deployment layouts, from a minimal single-node installation to a highly available cluster and a distributed architecture.
This is not a performance benchmark. Actual performance depends on hardware specifications, data distribution, and application access patterns. Without a defined workload, performance numbers are not meaningful.
Instead, this article compares three GBase Database deployment architectures from a structural perspective: where does each layout spend its resources, and what does it gain in return?
1. Three GBase Database Deployment Layouts at a Glance
The following comparison uses the same three-host infrastructure assumption where applicable.
| Dimension | Minimal Single Node | Single HA Group | Default Distributed |
|---|---|---|---|
| Hosts | 1 | 3 | 3 |
| gha_server | 1 | 2 | 1 |
| DCS | 1 | 3 | 3 |
| GTM | 1 primary | 1 primary + 1 standby | 1 primary + 1 standby |
| CN | 1 | 3 | 2 |
| DN groups / replicas | 1 / 1 | 1 / 3 | 2 / 3 |
| DN instances | 1 | 3 | 6 |
| Total listed instances | 5 | 13 | 14 |
| Average instances per host | 5 | ~4.3 | ~4.7 |
The instance averages are calculated from the listed component counts divided by the number of hosts. Actual per-host placement depends on the deployment topology.
The three layouts make different architectural trade-offs. Understanding these differences is more useful than comparing installation commands alone.
2. Resource Density: How Much Runs on Each Host?
The first visible difference is instance density.
The minimal single-node layout places five listed instances on one host. It has the lowest infrastructure entry cost, but all database components share the same machine resources.
The single HA group distributes its components across three hosts. It introduces additional DCS, GTM, CN, and DN instances to support availability and replication.
The default distributed layout runs six DN instances across two groups, together with CN, GTM, DCS, and gha_server components.
This increases the number of processes competing for memory, CPU, and I/O resources on each machine.
Its advantage is not necessarily better per-host efficiency. The benefit is that data can be distributed across multiple DN groups, providing a foundation for horizontal scalability.
Key takeaway: The minimal layout optimizes infrastructure entry cost, while the distributed layout allocates more resources to data distribution and expansion.
3. Failure Domains: What Happens When a Host Goes Down?
The second difference is availability and failure isolation.
Minimal Single-Node Deployment
The minimal layout has one DN instance and no DN replicas.
If a critical database process or the host fails, database service availability is affected. There is no replica within this layout to provide local database redundancy.
This makes it suitable for environments where low deployment cost and simplicity take priority over high availability.
Single HA Group
The single HA group has three DN replicas distributed across three hosts.
In the example topology, two replicas are placed in az1 and one in az2.
This arrangement can tolerate a single-host failure while retaining a majority. However, losing az1 entirely would remove two of the three replicas and therefore lose the majority.
This illustrates an important principle: replica count alone does not determine failure tolerance. Replica placement across hosts and availability zones matters just as much.
Default Distributed Deployment
The default distributed layout has two DN groups, each with three replicas.
Its two-group architecture allows data to be distributed rather than concentrated in a single group.
If the primary replicas are placed on different hosts, the impact of an individual host failure can be better isolated.
However, actual failure tolerance depends on replica placement, synchronization state, quorum requirements, and the failure scenario. Three replicas per group do not automatically guarantee availability under every multi-host or availability-zone failure.
Key takeaway: When evaluating GBase Database high availability, examine the actual replica placement and failure domains, not just the total number of replicas.
4. Scalability: Vertical Growth or Horizontal Expansion?
The third difference is how each architecture grows.
Minimal Single-Node Layout
The primary scaling approach is vertical: add CPU, memory, and storage resources to the existing host.
This is straightforward but ultimately constrained by the machine's capacity.
Single HA Group
The single HA group provides an availability-oriented architecture.
It can accommodate changes such as adding CN instances or adjusting the replica configuration, subject to supported topology operations and resource limits.
However, it remains a single DN group. Increasing resources or adding instances does not automatically create data distribution across multiple groups.
Default Distributed Layout
The default distributed architecture starts with two DN groups.
It provides a structural foundation for horizontal expansion by adding additional DN groups, such as dn3 and dn4, when supported by the deployment procedure and cluster configuration.
This is the key architectural distinction.
If future requirements include distributing data across additional groups, that decision should influence the initial deployment design.
Key takeaway: The minimal layout favors vertical scaling, the single HA group emphasizes availability, and the distributed layout provides a foundation for horizontal data expansion.
5. Port Planning: The Operational Cost of Each Layout
The fourth difference becomes visible in the port allocation table.
Different deployment layouts require different service ports and network access arrangements.
| Deployment Layout | Main Ports |
|---|---|
| Minimal single node | 20001, 2379, 6666, 5432, 15432 |
| Single HA group | Above ports plus 6001, 7001, 17001 |
| Default distributed | DN-related ports including 15432, 15433, 20010 |
The minimal layout has fewer port requirements and a simpler network configuration.
The single HA group introduces additional proxy-related ports, increasing the number of firewall rules and connectivity checks.
The default distributed layout requires additional DN-related port planning to accommodate its multiple groups and instances.
More ports mean more pre-installation verification, more firewall rules to maintain, and a higher possibility of port conflicts.
For GBase Database deployment teams, port planning should therefore be part of architecture design rather than something postponed until installation.
Always verify the complete port requirements against the official documentation for the target version and deployment mode.
6. Which GBase Database Architecture Should You Choose?
The three layouts serve different requirements.
| Requirement | Recommended Starting Point |
|---|---|
| Lowest infrastructure entry cost | Minimal single node |
| Database availability through replication | Single HA group |
| Data distribution across multiple DN groups | Default distributed |
| Primarily vertical expansion | Minimal single node |
| Future horizontal expansion | Default distributed |
Before choosing a GBase Database deployment architecture, ask three questions:
1. Can your application accept a single point of failure?
If not, a minimal single-node deployment is unlikely to satisfy your availability requirements.
2. Does your workload require data distribution across multiple DN groups?
If the answer is yes, a single HA group may not provide the architecture you need.
3. Will you need to add DN groups within the next two years?
If future horizontal expansion is a realistic requirement, plan for it before the initial deployment.
These questions often narrow down the architecture before detailed hardware sizing begins.
7. The Real Cost Is Decided Before Installation
In one sentence:
- Minimal single node buys the lowest infrastructure entry cost.
- Single HA group buys database availability through replication.
- Default distributed buys a foundation for horizontal scalability.
GBase Database(GBase 8c) provides different deployment options, but choosing between them is fundamentally an architecture decision.
The cost of switching later is not limited to installation effort. It can involve topology changes, replica placement, port planning, data distribution, and operational procedures.
That is why deployment architecture should be decided according to availability requirements, data distribution needs, and expected growth — not simply the easiest configuration to install today.
The best deployment layout is not the one with the most instances. It is the one whose architecture matches the system's actual requirements.
Top comments (0)