Architecture

The open-source KVM virtualization solution is designed for offline virtualization scenarios, including single-node, two-node high availability (HA), and multi-node cluster scenarios. It leverages VM migration and HA to ensure service reliability. Typical applications include databases, web servers, and cache servers.

• Single-Node Scenario Analysis

  In a single-node system, the QEMU-KVM open-source software is deployed on a single server. The Virt-Manager management software and virsh commands are used for out-of-band VM management. Both of them invoke libvirt APIs. The VNC software is used for in-band guest OS management. The local LVM virtual storage pool is used for VM storage. The VM network uses bridges (bridge mode) or physical NICs (host-only mode).

• Two-Node and Cluster Scenario Analysis

  The configurations of the two-node and cluster scenarios are based on that of the single-node system scenario. The computing virtualization, storage, and network configurations are the same as those of the single-node scenario. The difference is that the HA or live migration technology can be used to ensure cluster robustness in the two-node cluster and cluster scenarios. The following uses the Keepalived+LVS+MySQL two-node primary-secondary architecture as an example. Keepalived provides a floating IP address and periodically checks the health status of servers in the cluster. When a faulty node is detected, a switchover is triggered.

The architecture of the open-source KVM virtualization scenario is divided into three layers. The underlying layer is the Kunpeng server hardware, and the middle layer is the host Linux kernel and the KVM virtualization software. The top layer is the QEMU, which virtualizes I/O devices. Figure 1 shows the detailed system architecture and Table 1 describes the components.

Figure 1 Open-source KVM virtualization architecture