TencentOS Server is an enterprise-grade Linux server OS developed by Tencent Cloud, designed for CVM, bare metal servers, and container environments. The product is built on the community LTS kernel and optimized for cloud scenarios in areas such as networking, storage, resource scheduling, and container isolation.
TencentOS Server has been widely adopted within Tencent, covering businesses such as social networking, gaming, financial payments, AI, and security, and accounting for over 99% of Tencent's internal server OS deployments. It has been deployed on Tencent Cloud more than 10 million times cumulatively, managing over 50 million CPU cores.
Key Strengths
Extended support lifecycle: Each major version provides a maintenance lifecycle of up to 13 years, covering security updates, bug fixes, and vulnerability announcements to reduce migration costs caused by distribution end-of-life.
Smooth migration from CentOS: TencentOS Server 3 is fully compatible with CentOS 8 in user space, and TencentOS Server 2 is fully compatible with CentOS 7 in user space. Existing applications can typically be migrated without code modifications.
Kernel optimization for cloud scenarios: The self-developed Tkernel has been specifically optimized in subsystems such as CPU scheduling, memory management, I/O, and networking, with reproducible test data demonstrating performance improvements in certain scenarios.
Cloud native capabilities: provides container resource view isolation, container-level monitoring metrics, enhanced resource utilization, and eBPF support, and is compatible with mainstream orchestration platforms such as Kubernetes.
Hardware platform coverage: supports both x86_64 and ARM64 architectures, is compatible with processors from vendors such as Intel, AMD, NVIDIA, and Ampere, and provides multiple image formats including ISO, QCOW2, and Docker.
Product Architecture
TencentOS Server adopts a layered architecture design, which is divided from bottom to top into the hardware adaptation layer, kernel optimization layer, cloud native and resource management layer, and application and ecosystem layer. The responsibilities of each layer are shown in the following table:
|
Hardware adaptation layer | Adapts to x86_64 and ARM64 architectures, is compatible with mainstream processor platforms, and supports ISO, QCOW2, Docker, and other image formats. |
Kernel optimization layer | Built based on the community LTS kernel, with optimizations in subsystems such as CPU, memory, I/O, network, and eBPF. |
Cloud native and resource management layer | Provides capabilities such as container resource view isolation, container-level monitoring, resource utilization enhancement, and page cache isolation. |
Application and ecosystem layer | Compatible with mainstream open source software and Tencent self-developed products, and supports traditional enterprise applications and cloud native applications. |
Kernel Optimization
Tkernel has been optimized in the following subsystems, with corresponding measured data provided:
|
CPU | Optimize the CFS scheduler to improve scheduling fairness and flexibility and reduce process switching overhead. | Concurrent random number read performance improves by 28% compared with the community edition (based on stress-ng tests). |
Memory | Manage physical memory through the buddy system to reduce memory fragmentation, and use per-memcg lru lock to optimize concurrent access in multi-container scenarios. | Concurrent page cache access performance with multiple containers improves by 187% compared with the community edition. |
I/O | Supports multiple scheduling algorithms such as mq-deadline, bfq, and kyber, selectable based on business scenarios. | Reduce I/O latency and improve throughput. |
Network | Adopt the NAPI/POLL mechanism to handle high-traffic scenarios and prevent interrupt storms. | Improve network forwarding performance in high-concurrency scenarios. |
eBPF | Provides kernel-level programmability, allowing kernel features to be extended without modifying kernel source code. | Cover performance analysis, security audit, and network observation scenarios. |
Mirror version
Tencent Cloud provides three TencentOS Server images. The kernel, compatibility, and maintenance lifecycle of each version are shown in the following table:
|
TencentOS Server 4(TK5) | Linux 6.6 LTS | Independent technical roadmap, evolving based on the upstream community | Up to 13 years |
TencentOS Server 3(TK4) | Linux 5.4 LTS | Fully compatible with CentOS 8 | Up to 13 years |
TencentOS Server 2(TK4) | Linux 5.4 LTS | Fully compatible with CentOS 7 | Up to 13 years |
Version selection recommendations:
New business: Select TencentOS Server 4, which uses the community 6.6 LTS kernel and an independently evolved technology roadmap.
Migration from CentOS 8: Select TencentOS Server 3, whose user space is fully compatible with CentOS 8.
Migration from CentOS 7: Select TencentOS Server 2, whose user space is fully compatible with CentOS 7.
Use Cases
Containers and Cloud Native Platform
It is optimized for container scenarios in terms of resource isolation and memory management, and is compatible with container runtimes and orchestration platforms such as Docker, Containerd, and Kubernetes. It can serve as the OS foundation for microservices, Serverless, and Service Mesh architectures.
Internet Business
The TCP protocol stack and interrupt handling mechanism are optimized for high-concurrency network scenarios, making it suitable for network performance-sensitive services such as Web services, game servers, and audio/video services.
Enterprise Application Migration
For enterprise applications running on CentOS 7 or CentOS 8, you can select a compatible TencentOS Server version based on the current version to complete the migration and obtain long-term maintenance support.
AI and High-Performance Computing
It supports NVIDIA GPU and ARM64 computing platforms, with optimized I/O paths and memory management for AI scenarios. It is suitable for machine learning training, inference, and high-performance computing workloads.
Compatibility Description
Hardware Compatibility
It supports x86_64 and ARM64 architectures and is compatible with processors from vendors such as Intel, AMD, NVIDIA, and Ampere, with adaptations made for different hardware platforms.
Software Compatibility
Open-source services: Compatible with common services such as Apache, Nginx, Tomcat, Samba, FTP, NFS, PostgreSQL, and MariaDB.
Tencent self-developed products: Compatibility verification with TDSQL, Tencent Cloud-native Suite (TCS), and Tencent Cloud Enterprise (TCE) has been completed.
Container platform: natively supports container runtimes and orchestration platforms such as Docker, Containerd, and Kubernetes.
Services and Support
Tencent Cloud provides up to 13 years of maintenance and updates for each major TencentOS Server version, including security updates and vulnerability fixes. Existing servers can be patched through yum upgrades.
Note:
To learn more about TencentOS Server, you can consult Tencent Cloud Assistant through the mini program.