Today, I want to discuss an “old” topic—server virtualization, which is my area of expertise.
It is considered old because the core idea of this technology was born over half a century ago. However, it is also new because even today, it remains the cornerstone of our entire digital world. The apps on your phone, the OA systems in your company, and the cloud computing we use every day all rely on virtualization as the “unsung hero” behind the scenes.
However, many people, especially those who are new to the field, often feel confused when faced with terms like VMware, KVM, and Xen, as well as products like Huawei FusionCompute, H3C CAS, and Citrix. What is their relationship? What are the technical differences? Who will dominate the future?
Today, I will try to explain the “Three Kingdoms” of virtualization in simple terms, using some examples.
The Beginning: Why Virtualize?
Let’s rewind time to about 10 to 20 years ago. What were enterprise data centers like back then? Rows of cabinets with servers buzzing away. One server would run one application, for example, one server for the financial system, another for the email system, and yet another for the website.
What problems did this create?
- Resource Waste: The financial system might only be busy at the end of the month, with the server CPU utilization at only 5% most of the time; the email server might be idle at night, wasting CPU resources. A lot of computing resources are wasted.
- Operational Nightmare: With a variety of hardware brands and different system versions, administrators spend their days either installing systems or fixing hardware. If one application has a problem, it might require reinstalling the entire server.
- Space and Electricity Costs: As machines increase, data center space becomes scarce, and electricity bills rise sharply.
Virtualization technology was born to solve this “one carrot, one pit” dilemma. Its core idea is simple:to add a “middleman”—the Hypervisor (virtual machine monitor)—between the physical server and the operating system.
This “middleman” packages the CPU, memory, hard disk, network card, and other resources of a physical server into independent, virtual “small computers” (commonly known as virtual machines, VMs). You can run several virtual machines on this physical machine simultaneously, each with its own independent operating system and applications, isolated from each other.
This is like having a large three-bedroom apartment (the physical server) that used to be occupied by one person (one application), but now, through a middleman (Hypervisor), you rent out the three rooms to three different tenants (virtual machines), allowing them to live independently without disturbing each other. The utilization of the apartment has instantly increased.
The three main characters in our story today—VMware, Xen, and KVM—are currently the three most mainstream Hypervisor technologies.
The Differences in the ‘Martial Arts’ of the Three Major Technologies
Although their goal is “virtualization,” the implementation methods and “internal martial arts” of these three factions are vastly different.
1. VMware: The ‘Martial Arts Alliance Leader’
- Representative Product: VMware ESXi (the core of the vSphere suite)
- Position in the Industry: The pioneer of commercialization and once the absolute leader.
VMware can be considered the “Apple” of the virtualization field. When x86 server virtualization was still in chaos, it introduced a highly mature, stable, and user-friendly commercial solution.
Technical Characteristics:
Its Hypervisor, ESXi, is a highly optimized independent “small operating system” designed specifically for virtualization. It is installed directly on bare metal (the physical server) and does not rely on any other operating system. This architecture is referred to as “Type 1 – Bare Metal“.
To understand this, let me give an analogy:
VMware ESXi is like an extremely experienced “five-star hotel manager.” This manager is a complete management system in itself, not needing to rely on others, directly managing all the resources of the hotel building (physical server) and accurately allocating them to each room’s guests (virtual machines), providing a full suite of top-notch services from check-in, utilities, security to room service (vMotion, HA, DRS, and other advanced features).
Advantages: Excellent performance, stable and reliable, with a highly complete ecosystem. From virtualization, storage (vSAN) to networking (NSX), VMware provides a “full package” service, making it easy to resolve issues by contacting just one company.Disadvantages: Expensive! Very expensive! Moreover, it is technically closed, typical of commercial software, making it easy to be “vendor-locked”.
2. Xen: Once a Genius, Now a ‘Specialized Student’
- Representative Product: Citrix Hypervisor (formerly known as XenServer), AWS was also initially based on Xen.
- Position in the Industry: A pioneer in open-source virtualization, once a strong competitor to VMware.
Xen originated from the Cambridge University lab and is a darling of the open-source community. It was the first to propose the concept of “Paravirtualization (PV)“, which is a very clever idea.
Technical Characteristics:
Xen’s architecture includes a special virtual machine called “Domain 0” (privileged domain). This Dom0 typically runs on a minimal Linux system and is responsible for managing hardware drivers and controlling all other virtual machines (Domain U).
- Paravirtualization (PV): It assumes that the operating system in the virtual machine (Guest OS) is “aware” that it is running in a virtual environment. By modifying the operating system kernel, it allows the OS to cooperate with the Hypervisor, thereby reducing the performance loss caused by virtualization. This is like a driver (Guest OS) knowing they are driving a “training car” and actively cooperating with the instructor (Hypervisor), making the driving experience very efficient.
- Full Virtualization (HVM): Later, with the popularization of hardware virtualization technologies (Intel VT-x, AMD-V), Xen also supported full virtualization without needing to modify the system kernel.
To draw a comparison:
Xen is like a “general contractor (Dom0)” leading a group of “subcontractors (Guest VMs).” This general contractor first settles on the construction site (physical server), mastering all the blueprints and tools (hardware drivers), and then directs the subcontractors to do their work. This model was very efficient in the early days, but the stability and performance of the general contractor became critical.
Current Analysis:
Although Xen’s architecture is clever, the existence of Dom0 makes it relatively complex. In the wave of cloud computing, it has gradually been overshadowed by the later entrant KVM. Today, its most successful application scenario remains in Citrix’s “old business”—desktop virtualization (VDI), where it still holds significant power.
3. KVM: The ‘Chosen One’ of the Linux Kernel
- Representative Product: All mainstream Linux distributions come with KVM built-in, and domestic products like Huawei FusionCompute, H3C CAS, Alibaba Cloud, Tencent Cloud, etc., are almost all based on KVM with deep customizations.
- Position in the Industry: The king of open-source virtualization, the de facto standard in the cloud era.
The emergence of KVM (Kernel-based Virtual Machine) can be seen as a “dimensionality reduction attack” in the virtualization field. Its approach is extremely “brutal” yet effective:instead of reinventing the wheel, it directly turns the powerful Linux kernel into a Hypervisor.
Technical Characteristics:
KVM itself is not a complete Hypervisor; it is merely a module of the Linux kernel. Once this module is loaded, your Linux system transforms into a “Type 1 – Bare Metal” Hypervisor.
- It cleverly reuses the decades of achievements accumulated by the Linux kernel: process scheduling, memory management, device drivers, security models, etc.
- Each virtual machine, from the perspective of the Linux host, is just a regular Linux process, visible with the `ps` command and can be “killed” with the `kill` command.
To use another analogy:
KVM no longer needs an external “hotel manager” or “general contractor.” It directly embeds management capabilities into the building (Linux kernel) itself. The building inherently possesses the ability to allocate resources, ensure security isolation, and manage visitors. Each virtual machine (tenant) is an “independent room (process)” managed uniformly by the building. This “native” management approach is naturally the most direct and efficient.
Advantages: Open-source, free, powerful performance, and seamless integration with the Linux ecosystem. Being part of the kernel, it is very lightweight and stable.Disadvantages: The native KVM management tools are relatively rudimentary, requiring additional components like QEMU and libvirt to form a complete solution, which presents a certain learning curve. However, this also provides ample space for various vendors to conduct secondary development and productization.
The Future Trends of Virtualization in China
After discussing the technology, let’s shift our focus back to the domestic market and examine the future directions of these three factions.
-
VMware’s ‘Twilight’: VMware still holds an important position in the domestic high-end enterprise market, especially in finance and foreign enterprises, due to its years of stability and strong ecosystem. However, its “good days” are facing significant challenges. Firstly, there is the cost issue; high licensing fees deter many enterprises. Secondly, and most critically, under the current backdrop of “self-control” and “trusted innovation,” its closed commercial model and “foreign” identity are gradually marginalizing it in key areas like government and state-owned enterprises.It is foreseeable that VMware’s market share in China will continue to shrink, gradually retreating to specific markets that are extremely sensitive to migration costs.
-
Xen/Citrix’s ‘Perseverance’: Xen’s presence in the general server virtualization market in China has diminished. Its main battlefield, or almost its only battlefield, is desktop virtualization (VDI). Citrix remains a prestigious name in this field, and many enterprises still think of it first when they need remote work and secure desktop solutions.In the future, Xen/Citrix will continue to serve as a “specialized expert,” delving deep into the VDI niche but will find it difficult to impact the mainstream virtualization landscape.
-
KVM’s ‘Dominance’:There is no doubt that the KVM camp is the future of the domestic virtualization market. There are three reasons:
- Policy Favor: KVM is open-source, perfectly aligning with the national strategy of “self-control.” Domestic vendors can access the source code for deep customization and security reinforcement, creating truly proprietary products.
- Cloud-Native Wave: KVM serves as the foundation of the entire cloud-native ecosystem. Whether it is OpenStack private cloud or container technologies (like Kata Containers, which rely on KVM), they are closely tied to KVM. The future is the era of the cloud, and it is also the era of KVM.
- Rising Ecosystem: Domestic first-tier vendors, represented by Huawei FusionCompute and H3C CAS, have already developed mature products based on KVM that are “comparable to” or even “surpassing” VMware in functionality and experience (though there are still some performance gaps). They not only provide virtualization platforms but also integrate storage, networking, and cloud management capabilities, along with more competitive pricing and localized services, rapidly capturing market share from VMware.
In Conclusion
The virtualization landscape has evolved from VMware’s dominance to Xen’s strong challenge, and now to KVM’s unification, reflecting the entire IT industry’s shift from closed to open, from traditional IT to cloud computing.
For us technical professionals, understanding the evolution and differences of these technologies is not just to cope with work but also to see the trajectory of technological development. VMware’s success teaches us the importance of productization and ecosystem; Xen’s ingenuity showcases another possibility for technological breakthroughs; and KVM’s victory confirms that embracing open-source and integrating into the mainstream ecosystem is the way forward.
In the future, virtualization technology itself may become increasingly “transparent,” and we will discuss more about upper-layer applications like containers and Serverless. However, regardless of how the upper structures change, KVM, as a solid foundation, will continue to silently support our ever-expanding digital world for a long time to come.