20.2.1 Network Takeover

PRIMECLUSTER provides architectural patterns for taking over the network in cluster systems on public clouds. For smooth design of a cluster system, choose from these architectural patterns.

The following are the architectural patterns for taking over the network and the appropriate scenarios for each pattern.

Table 20.1 Architectural patterns and appropriate scenarios for taking over the network
Architectural pattern	Appropriate scenario	Note
Network takeover by the virtual router	The cluster system of back-end servers used by the client in the VPC	The cluster system is secured as an access from public sites is blocked. The client must be deployed in the VPC.
Network takeover by replacing the Elastic IP address	The cluster system of front-end servers	Accessible from public sites. To assure the security, additional access control is required.
Network takeover by rewriting DNS records	The cluster system of back-end servers used by the client on-premises or in any other VPC	Accessible via VPN from the client. An additional device is required for a VPN connection.

20.2.1.1 Network Takeover by the Virtual Router

This is an architectural pattern using the cluster system from the client in the VPC. Select this architecture when building the cluster system in back-end servers used by the client in the VPC.

In this architectural pattern, the network takeover is enabled by rewriting the route table of the virtual router provided by AWS. In the event of a cluster node error, PRIMECLUSTER automatically rewrites the route table and switches to the standby system to take over the IP.

An access from public sites (clients outside of the VPC) is blocked and the cluster system is secured.

Figure 20.1 Taking over the network by the virtual router

20.2.1.2 Network Takeover by Replacing the Elastic IP Address

This is an architectural pattern that allows an access to the cluster node from public networks (the Internet). Select this architecture when building the cluster system in front-end servers.

In this architectural pattern, the network takeover is enabled by associating and controlling the operational network interface as a transfer destination for the Elastic IP obtained as the takeover IP. In the event of a cluster node error, PRIMECLUSTER changes the transfer destination of the takeover IP to the network interface of the takeover destination of the application and switches to the standby system to take over the IP.

The cluster system provides the client with the Elastic IP address as the takeover IP. The client accesses this Elastic IP address.

Note

To prevent unauthorized accesses from public sites with the administrator authority, do not co-use the administrative LAN and the public LAN.

Figure 20.2 Taking over the network by replacing the Elastic IP address

20.2.1.3 Network Takeover by Rewriting DNS Records

This is an architectural pattern using a VPN service from the client system on-premises to allow a direct access to the cluster node. Hybrid configurations such as on-premises for the application layer and cloud for the database access layer are possible. Select this architecture if the network takeover by rewriting the route table is unavailable when using a VPN service.

In this architectural pattern, the network takeover is enabled by rewriting records of a DNS service provided by AWS. In the event of a cluster node error, PRIMECLUSTER takes over the IP by replacing the related records held by a DNS service with the standby private IP.

Figure 20.3 Taking over the network by rewriting DNS records