Security and Stability Validated Across Over 10 Million Nodes
Tencent has a complete operations management system for TencentOS Server and has designed an automatic crash analysis system. It can detect anomalies in overall system operations in real time, proactively collect information about OS exceptions, analyze and identify anomalies, and fix faults in real time through hot patches.
After more than 10 years of continuous refinement, TencentOS Server had been deployed on over 10 million nodes cumulatively in Tencent's internal and customer environments as of July 2026, with operational practices accumulated across multiple business scenarios.
Note: The data of 10 million nodes is based on cumulative deployment statistics in Tencent internal and customer environments as of July 2026.
Comprehensive Optimization and Higher Performance
TencentOS Server is a deeply optimized high-performance OS with optimizations applied to various types of software within the system. In specific business scenarios such as high-concurrency network services and big data storage and access, typical business performance is improved by more than 50%. TencentOS Server 4 adopts the self-developed TGCC12 compiler (based on GCC 12.3.1), achieving a performance improvement of more than 10% in compilation optimization and runtime efficiency compared with TencentOS Server 3.
The above performance data is based on Tencent internal experimental test results in 2026. For details on the test environment, scenarios, and comparison versions, refer to the respective feature descriptions. The data is for reference only and does not represent actual performance in all scenarios. Actual performance may vary depending on the usage environment.
For CVM and bare metal scenarios, TencentOS Server provides deeply optimized networking, storage, and resource isolation features such as cloud-native scheduler, RUE resource isolation, and high-performance network protocol stack. It supports high-bandwidth and high packet throughput scenarios, and provides resource isolation and security enhancement capabilities. Actual results depend on the configuration and business scenario.
User-Space Compatibility
TencentOS Server 2 and 3 use a kernel developed in-house, while their user-space components share the same upstream sources as CentOS and maintain compatibility with the corresponding CentOS components. These components originate from the same upstream open-source software repositories and are rebuilt and repackaged into a distribution. Compatibility support is provided within the supported CentOS versions and user-space software ecosystem, and applications originally running on CentOS can typically run directly on TencentOS Server. The specific compatibility scope is subject to the compatibility list and migration test results. It is recommended to complete business validation and canary testing before migration.
Additionally, TencentOS Server provides migration tools that support in-place migration of CentOS in scenarios that meet product requirements. Actual migration results are affected by factors such as the source system version, software dependencies, drivers, and business configurations. It is recommended to complete compatibility assessment and testing before migration. These tools have been applied in various scenarios within Tencent.
TencentOS Server 4 is backed by the OpenCloudOS community, with its kernel and user-space software independently evolved based on the upstream community. The specific software package sources and maintenance methods are subject to the technical white paper.
In-House Features for Cloud-Native Scenarios
TencentOS Server leverages Tencent's continuous advancements in the cloud-native field to develop a comprehensive full-stack open-source cloud-native solution. As the foundation at the bottom of the entire technology stack, it provides core support for the stable and efficient operation of the full cloud-native stack.
At the kernel level, TencentOS Server has developed and integrated a complete set of original cloud-native features to form a cloud-native kernel. Key components include Tencent Cloud Native Scheduler (TCNS), RUE for resource isolation, service quality monitor, cloud-native SLI (specialized container-level metrics), Mbuf for persistent monitoring of critical kernel resources, and cgroupfs for container resource view isolation. Typical features are described as follows:
The cloud-native resource isolation system (RUE) supports resource isolation in multi-priority hybrid deployment scenarios.
Cloud-native SLI provides a more refined and professional monitoring system and metrics for cloud-native workloads.
Quality monitor and Mbuf provide more timely and continuous tracing capabilities for cloud-native workloads.
cgroupfs provides an independent resource view for cloud-native workloads.
It supports the eBPF capabilities adapted for the product in Linux 6.6 LTS, as well as some key features of Cilium. For details, refer to the compatibility list.
NetTrace is a full lifecycle tracking system for network packets in complex cloud-native network scenarios, enhancing network maintenance capabilities in cloud-native environments.
Comprehensive Support for AI Scenarios
TencentOS Server continues to enhance its AI scenario support. As of July 2026, it is compatible with over 40 GPU/NPU hardware models, supports 30+ mainstream AI frameworks, and covers scenarios such as training and inference.
The above data is based on Tencent internal compatibility tests and the product support list as of July 2026. For specific hardware models, driver versions, framework versions, and architecture coverage, refer to the compatibility list.
AI-enhanced public image: An AI-Ready image built on TencentOS Server 4, preinstalled with core components such as CodeBuddy CLI, Node.js, Docker, and pip3, along with expert operations skills for TencentOS. It is designed for two major scenarios: AI Agent deployment and AI-assisted operations. The preinstalled components can reduce some environment configuration work, but actual deployment still requires validation based on hardware models, driver versions, application dependencies, and resource configurations.
AI container images: An out-of-the-box set of container images with AI runtime dependencies such as NVIDIA Driver, CUDA, and cuDNN preinstalled, reducing some environment configuration work. Actual deployment still requires validation based on GPU/NPU models, driver versions, application dependencies, and resource configurations.
Security Compliance and Hot Patching
TencentOS Server leverages its self-developed vulnerability management platform, with third-party intelligence platforms and YunDing Lab as information sources. After obtaining vulnerability intelligence, it quickly analyzes and reproduces the vulnerabilities, performs targeted fix verification, and ultimately delivers fixes to users through hot patches. Meanwhile, the Tencent OS team has established clear SLA-based reporting mechanisms for security vulnerabilities of different severity levels, reducing fix time and avoiding unnecessary downtime for users. For details, see TencentOS Server Security Advisory. Professional Maintenance and Continuous Support
In terms of technical service system, Tencent has a relatively complete technical service system and extensive, long-term technical service experience for large-scale server scenarios. Specific response times, service scopes, applicable packages, exclusions, and service levels are subject to the service agreement or SLA. In terms of technical service team, Tencent uses a self-developed kernel, began independent OS development in 2010, has a large number of kernel experts, and has accumulated experience in continuously contributing to the upstream community, as well as strong technical influence.
Note:
Third-party names, trademarks, and software mentioned in this document (including but not limited to CentOS, GCC, Linux, OpenCloudOS, Cilium, NVIDIA, CUDA, cuDNN, Node.js, and Docker) belong to their respective owners. They are used only to indicate compatibility or adaptation scope and do not constitute endorsement, warranty, or recommendation.