Microsoft Hyper‑V Architecture Overview

Microsoft acquired Connectix, the creator of Virtual PC, in 2003, marking its entry into virtualization and later releasing Virtual Server. Facing dominant competitors such as VMware and Xen, Microsoft kept Virtual PC and Virtual Server low‑profile, but the launch of Hyper‑V brought the technology into the spotlight and reshaped the market.

Microsoft Hyper‑V Architecture

Compared with the more familiar VMware architecture, Hyper‑V is also a server‑virtualization hypervisor (similar to VMware ESXi). Microsoft added Hyper‑V support starting with Windows Server 2008, virtualizing server compute resources and managing VMs through Hyper‑V Manager; cluster management is handled by System Center VMM.

Microsoft offers several products and deployment options for Hyper‑V:

Free Windows Hyper‑V Server – a lightweight version for virtualization without clustering or advanced features.

Full Windows Server installation with the Hyper‑V role enabled; Hyper‑V functions as a system module that can be turned on or off.

Server Core installation mode, which removes the GUI and other non‑essential components to improve reliability.

Virtual Server and Virtual PC are based on a hybrid virtualization model where the VMM (hypervisor) and the host server run at the same kernel level and alternate access to CPU resources; VMware Workstation, by contrast, runs the VMM as an application on top of the operating system.

Hyper‑V Architecture

In Windows Server 8 beta and Windows 2012 R2, Hyper‑V is fully 64‑bit and requires the processor to support AMD‑V or Intel VT hardware‑assisted virtualization. Its architecture has several distinctive components.

Parent partition – part of the Hyper‑V virtualization program that provides services to child partitions.

VSP (Virtualization Service Provider) – directly interacts with each hardware device and offers hardware and file‑system services.

VSC (Virtual Service Consumer) – the client component inside a child partition that consumes services provided by VSP.

VMBus – the communication channel between VSP and VSC; each hardware device has a corresponding VSP/VSC pair.

Microsoft Early Product Architecture

Earlier products such as Virtual Server 2005 R2 and VMware did not require processor‑assisted virtualization (VT/AMD‑V) as a mandatory feature, unlike Hyper‑V where it is compulsory.

Hyper‑V Microkernel Architecture

Hyper‑V employs a microkernel design where the most essential functions (process scheduling, memory management, inter‑process communication) run in kernel mode (Ring 0) while most other services run as separate user‑mode processes (Ring 3). Communication between these components uses IPC mechanisms, providing a lightweight and secure architecture.

Most Linux systems (and other Unix‑like OSes) use a monolithic kernel where all modules reside in a single large binary and communicate via direct function calls, which can offer performance advantages over a microkernel.

Hyper‑V Networking

Hyper‑V virtual networking connects VMs through VMBus and a virtual switch (vSwitch), implementing three types of virtual networks:

Private virtual network – allows communication only among VMs on the same Hyper‑V host; VMs cannot reach the host.

Internal virtual network – permits VM‑to‑VM communication and VM‑to‑host communication, but no external network access.

External virtual network – enables communication among VMs, between VMs and the Hyper‑V host, and with external networks.

Hyper‑V vSwitch

The vSwitch provides features such as multi‑tenant isolation, VM queue (VMQ) for traffic shaping, extensible SDN capabilities (OpenFlow, NIC teaming), and integration with NDIS and Windows Filtering Platform drivers to enforce security and compliance for non‑Microsoft extensions.

SR‑IOV (Hyper‑V Single Root I/O Virtualization)

SR‑IOV maps portions of a physical NIC directly to a VM, allowing the VM to access the NIC without passing through the vSwitch, thereby reducing latency and CPU overhead.

VMQ (VM Queue)

VMQ allocates a dedicated receive queue on the physical NIC for each VM. When the NIC receives data, it places the packets directly into the corresponding VM’s queue, allowing the VM to process data without the vSwitch copying it.

Warm Tip: Please search “ICT_Architect” or scan the QR code below to follow the public account and get more exciting content.