MULTI-POD SOLUTION DESIGN

This section covers the detailed design of the multi-PoD solution from the following aspects: application scenario, specific design, and deployment design.

Application Scenario of the Multi-PoD Solution

In the multi-PoD solution, when multiple DCs are within a short distance from each other, a single SDN controller remotely manages multiple DCs, and multiple DCs are in the same resource pool. This way, a large Layer 2 network across DCs can be constructed, deploying a VPC across DCs. In addition, SDN controller clusters are deployed in active/standby mode, and DR and backup are performed on the management plane. The multi- PoD solution supports cross-DC deployment of clusters and migration of VMs, as well as network-level active/standby DR.

Logically, network features and behaviors in the multi-PoD scenario are similar to those on a single-DC network. You only need to focus on special requirements in the multi-DC scenario. The following details the application scenarios.

Cross-DC Cluster Deployment

As cluster technologies mature, an increasing number of active-active DCs are deployed in a cluster to ensure cross-region availability and provide load balancing capabilities.

Physical IP addresses are used to provide services for database and application servers. This mode applies only to data applications (in client/ server mode). To improve service continuity, clusters can be deployed across DCs, in which case cluster servers are deployed in different DCs and provide unified access interfaces. In addition, service IP addresses are replaced by VIPs, and negotiation and status synchronization are performed between server clusters through the cross-DC interconnection network. The cluster heartbeat link and public network usually need to access the same Layer 2 domain. Therefore, a large Layer 2 network needs

Cross-DC cluster deployment

FIGURE 7.28 Cross-DC cluster deployment.

to be deployed across DCs. Layer 2 interconnection can be implemented using technologies such as bare optical fibers, Wavelength Division Multiplexing (WDM), VPLS, and EVPN VXLAN, as shown in Figure 7.28.

The server cluster solution has the following requirements on DCNs:

  • • Low latency: The round-trip time (RTT) and distance of deployment are limited.
  • • Layer 2 interconnection: Layer 2 interconnection is required, and the latency in interconnection is limited.
  • • High network reliability: The interconnection network is highly reliable to avoid split-brain occurring.
 
Source
< Prev   CONTENTS   Source   Next >