Deepin aBos Integrated Machine: Architecture and Virtualization Technologies

The aBos integrated machine solution combines networking, compute, and storage resources into a single appliance for small‑to‑medium enterprises or branch offices, featuring three core components—network device virtualization, server virtualization, and storage virtualization—managed through a web control platform (VMP) and a centralized BBC management center.

Server virtualization in aBos uses the aSV hypervisor, which presents virtual machines (VMs) to end users, allowing them to install software, attach disks, and configure networks just like on a physical x86 server.

Hypervisor Architecture

A hypervisor sits between physical hardware and operating systems, enabling multiple OS instances to share the same hardware. It allocates memory, CPU, network, and disk resources to each VM and loads the guest OS.

Hypervisors are classified into two types:

Type‑I (bare‑metal) runs directly on hardware without a host OS, offering better performance; examples include VMware ESX, Citrix XenServer, Microsoft Hyper‑V, and Linux KVM.

Type‑II (hosted) runs on top of a host OS, incurring additional overhead; examples include VMware Workstation, Virtual PC, and Virtual Server.

Deepin's aSV adopts a Type‑I approach using Linux KVM, which is a kernel‑based virtualization module leveraging hardware extensions (Intel VT‑x) and the QEMU user‑space tool.

KVM provides the kernel module, while QEMU supplies the user‑space emulation of hardware devices; together they enable full control of virtual machines.

vCPU Mechanism

Each virtual CPU (vCPU) is represented by a VMCS structure; when a vCPU is scheduled onto a physical CPU, its context is saved to or restored from this structure, allowing independent execution of multiple vCPUs.

Memory Virtualization

The VMM manages a two‑level address translation: Guest Virtual Address (VA) → Guest Physical Address (PA) → Machine Address (MA). Techniques such as Extended Page Tables (EPT) let the CPU perform both translations in hardware, eliminating the need for software‑based shadow page tables.

I/O Device Virtualization

Three main I/O virtualization methods are described: full device emulation, front‑end/back‑end split drivers, and direct device assignment (e.g., via IOMMU/PCI‑SR‑IOV), each balancing performance and compatibility.