Multi-PoD Solution Des ign

Architecture of the Multi-PoD Solution

Multi-PoD is mainly used in scenarios where DCs are deployed in the same city within a short distance, and where the network layer provides network DR.

In Figure 7.31, compute and network resources of multiple DCs form a unified resource pool and are managed by one SDN controller. VPCs and subnets can be deployed across DCs and can communicate with each other at Layer 2 or Layer 3.

To improve the reliability of the management plane, SDN controllers can be deployed in the two DCs. The two SDN controller clusters work in active/standby mode. The active SDN controller cluster manages the network. If the active SDN controller cluster is faulty, an active/standby switchover is performed between the two clusters, and the original standby SDN controller cluster becomes active to take over services, performing DR on the management plane.

In this scenario, the physical networks of each DC are independent of each other in terms of architecture. Devices and channels for interconnection between DCs need to be added on the basis of a single DC. Multiple DCs are unified end-to-end VXLAN domains, and the network and computing resources are unified resource pools.

It is recommended that active and standby egress gateways be deployed; that is, all north-south traffic is diverted to the active egress gateway. If the egress gateways work in active/standby mode, firewalls are also deployed in active/standby mode to ensure high service availability.

Figure 7.32 shows main features of the multi-PoD solution.

  • • DR on the management plane of a single domain: Multiple DCs share a resource pool. In a VXLAN domain, only one resource pool is available to services, and VPCs can be deployed across DCs. In addition, SDN controller clusters are deployed in active/standby mode, performing DR on the management plane.
  • • Cross-DC VPC deployment: Similar to the DR on the management plane of a single domain, a VPC can be deployed across fabric networks because there is one resource pool.
  • • Cross-DC deployment of egresses in active/standby mode within a VPC: Multiple DCs can have multiple egresses. One egress can be selected as the external gateway for a VPC, while the other egresses are used as standby egresses. The active and standby egresses are subsequently determined based on route priorities. You can also select only one centralized egress through which all north-south traffic is transmitted to northbound APIs.
  • • Cross-DC deployment of firewalls in active/standby mode in a VPC: Firewalls are deployed in active/standby mode on the fabric network, and two groups of firewalls are deployed between fabric networks. The controller delivers configurations and policies to the two groups of firewalls, within which the active and standby firewalls are determined by the active and standby routes.
  • • When the SDN controller cluster needs to connect to the OpenStack cloud platform (Huawei FusionSphere platform can be also used): One OpenStack can remotely manage multiple DCs. To ensure reliability, OpenStack can be deployed across DCs in a cluster. If the SDN controller needs to connect to the VMM, you are advised to deploy one VMM in each DC and connect one SDN controller cluster to multiple VMMs.
 
Source
< Prev   CONTENTS   Source   Next >