TencentOS Server is a domestic enterprise-grade Linux server OS launched by Tencent Cloud. As of July 2026, it has been deployed at a cumulative scale of tens of millions of nodes across Tencent's internal and customer environments. TencentOS Server demonstrates solid technical capabilities and strong performance in core scenarios such as system migration, database foundations, cloud native support, and LLM inference acceleration.
System Migration: Smooth Migration and Stable Upgrade
Scenario Description
As some mainstream Linux distributions gradually stop maintenance, many enterprises face severe challenges such as unfixable system security vulnerabilities and a lack of community support. Users in industries such as internet, finance, and government that have long relied on the existing Linux ecosystem urgently need a replacement OS with good compatibility, low migration costs, and greater stability.
Pain Point Analysis
Service Discontinuation Risk: After the original system stops updates, security patches are no longer provided, system vulnerabilities remain continuously exposed, and production environments face dual pressures of compliance and security.
High Migration Costs: Replacing traditional systems requires reinstalling environments and re-adapting applications, leading to prolonged business interruptions and persistently high labor and time costs.
Difficult Ecosystem Compatibility: Some alternative solutions are incompatible at the user space level, causing existing applications to fail to run properly and frequent failures after migration.
Solution
TencentOS Server evolves independently based on the community LTS kernel and maintains high compatibility with mainstream Linux distributions under the user space software, hardware architectures, and configuration conditions covered by the compatibility list. Before migration, you are advised to perform thorough business validation in a test environment to confirm the compatibility of application dependencies, kernel modules, drivers, and third-party software. A self-developed migration tool is also provided to support in-place migration of the original system, typically allowing OS replacement without reinstalling the system. The specific compatibility scope, migration steps, and precautions are subject to the migration guide, helping reduce migration costs and business interruption risks.
Benefit
|
Compatibility | Maintains high compatibility with mainstream Linux distributions within the scope covered by the compatibility list. |
Migration Method | Supports in-place migration and typically does not require system reinstallation. |
Service guarantee | The availability of related cloud services is subject to the applicable Service Level Agreement. |
While maintaining ecosystem compatibility with mainstream Linux distributions, TencentOS Server delivers stronger service capabilities and stability through deep kernel optimization, making it a viable migration option after the original system reaches end of service.
Note:
The above performance, efficiency, and energy consumption data comes from Tencent's internal experimental test results in 2026. The test environment and scenarios were Tencent's internal laboratories, covering typical database, cloud native, and AI inference workloads. The results are for reference only and do not represent actual performance in all scenarios. Actual results may vary depending on hardware specifications, business workloads, kernel versions, configurations, model frameworks, and runtime environments. Refer to actual usage conditions for accurate information.
The availability and service level commitments of cloud services related to TencentOS Server are governed by the applicable Service Level Agreement (SLA). For applicable services, statistical periods, exclusions, and compensation rules, refer to the corresponding SLA document.
TDSQL/TCE Infrastructure: Software-Hardware Synergy and Performance Optimization
Scenario Description
In enterprise core database and private cloud platform construction, the OS serves as the underlying infrastructure and directly affects database write performance, system response speed, and overall throughput. TencentOS Server forms a collaborative solution with Tencent Cloud's self-developed database TDSQL and the enterprise-grade cloud platform TCE, delivering a software-hardware coordinated OS foundation through combined optimization policies.
Pain Point Analysis
Database Write Bottleneck: Under high concurrency, file system I/O becomes a performance bottleneck, affecting business real-time performance.
Long Startup Warm-Up Time: After a database instance starts, it requires a long warm-up period to reach optimal performance, which affects fault recovery and elastic scaling efficiency.
Insufficient Overall Performance: General-purpose OSs are not tuned for database scenarios, so they cannot fully utilize hardware performance, resulting in resource waste.
Solution
TencentOS Server performs kernel-level joint tuning for TDSQL/TCE and implements optimization policies in file systems, memory management, and I/O scheduling. Through XFS file system write path optimization, a database warm-up acceleration mechanism, and overall performance tuning, it achieves deep collaboration between the OS, databases, and cloud platforms.
Benefit
|
Write performance | Compared with the unoptimized kernel, the file system (XFS) write performance is improved by more than 20%. |
Start time | Compared with the unoptimized kernel, the database warm-up time is reduced by more than 30%. |
Overall performance | Compared with the unoptimized kernel, the sysbench-measured performance is improved by 5% to 25%. |
The combined optimization solution of TencentOS Server and TDSQL/TCE can improve runtime performance in specific database and cloud platform scenarios through targeted tuning. Actual results depend on test conditions and real business environments. This solution is applicable to database and cloud platform scenarios such as TDSQL/TCE.
Note:
The above performance, efficiency, and energy consumption data comes from Tencent's internal experimental test results in 2026. The test environment and scenarios were Tencent's internal laboratories, covering typical database, cloud native, and AI inference workloads. The results are for reference only and do not represent actual performance in all scenarios. Actual results may vary depending on hardware specifications, business workloads, kernel versions, configurations, model frameworks, and runtime environments. Refer to actual usage conditions for accurate information.
Cloud Native Support: Resource Efficiency and Green Energy Saving
Scenario Description
In the cloud native era where containerization and microservices are widely adopted, enterprises face multiple challenges such as low CPU utilization, high memory costs, GPU resource contention, and rising data center energy consumption. TencentOS Server performs deep kernel-level optimization for cloud native scenarios, providing multi-dimensional resource optimization capabilities across compute, memory, and GPU.
Pain Point Analysis
Low CPU Utilization: Online and offline services are deployed independently, so resources cannot be scheduled in a unified manner, and average CPU utilization remains low over a long period.
High Memory Costs: Memory demand continues to grow, and without effective memory compression and reclamation mechanisms, memory avalanches can easily occur, leading to service interruptions.
GPU Resource Contention: AI training and inference tasks exclusively occupy GPUs, resulting in low resource utilization. When multiple tasks share a GPU, precise partitioning capabilities are lacking.
High Energy Consumption Pressure: Data center power costs continue to rise, and server energy consumption lacks intelligent adjustment mechanisms.
Solution
TencentOS Server provides four core cloud native capabilities:
Colocation of Online and Offline Services: Unify the scheduling of compute resources to significantly improve CPU utilization while ensuring high QoS.
Memory Compression Offloading: Through intelligent memory reclamation and a multi-level demotion mechanism, it saves more than 20% of memory compared to unoptimized kernels, helping to reduce memory pressure and exception risks.
GPU Partitioning: Supports multiple business containers sharing the same physical GPU to achieve precise partitioning of compute power and memory, improving GPU resource utilization by approximately 60% compared to non-partitioned scenarios.
System-Wide Energy Consumption Optimization: Through intelligent CPU power state adjustment and hard disk deep sleep management, power consumption is reduced by 5% to 30% compared to unoptimized systems, and nodes can be woken up within seconds.
Benefit
|
CPU utilization | Colocation of online and offline workloads significantly improves utilization. |
Memory Savings | Compared with the unoptimized kernel, memory compression offloading saves more than 20%. |
GPU utilization | Resource utilization after GPU partitioning increases by approximately 60% compared with no partitioning. |
Overall server power consumption | Power consumption reduced by 5% to 30% |
TencentOS Server helps enterprises reduce costs and improve efficiency with multi-dimensional resource optimization capabilities, making it suitable for cloud native scenarios that prioritize performance, cost, and ecosystem compatibility.
Note:
The above performance, efficiency, and energy consumption data comes from Tencent's internal experimental test results in 2026. The test environment and scenarios were Tencent's internal laboratories, covering typical database, cloud native, and AI inference workloads. The results are for reference only and do not represent actual performance in all scenarios. Actual results may vary depending on hardware specifications, business workloads, kernel versions, configurations, model frameworks, and runtime environments. Refer to actual usage conditions for accurate information.
AI infra: Adaptation, Deployment, and Ops Support for AI Scenarios
Scenario Description
As the demand for private deployment of AI large models on the enterprise side surges, users commonly face challenges such as complex environment configuration, difficult hardware adaptation, and high Ops complexity. TencentOS Server AI Enhanced Edition provides software and hardware adaptation, model deployment, and intelligent Ops capabilities for AI scenarios. It integrates mainstream inference frameworks and AI software stacks, covers supported hardware adaptation, model deployment, and Ops processes, helping enterprises build AI infrastructure. Pain Point Analysis
Complex Environment Configuration: Private deployment of large models involves multiple steps such as driver installation, framework configuration, and dependency management, making environment setup time-consuming and error-prone.
Difficult Hardware Adaptation: Enterprise GPU and NPU hardware comes in a wide variety of models (such as NVIDIA and AMD), making individual adaptation costly.
High Ops Complexity: Troubleshooting in AI inference environments is difficult, and issues such as kernel crashes and performance bottlenecks lack professional diagnostic tools.
Difficult Stability Assurance: Long-running inference tasks are susceptible to system-level issues such as OOM and scheduling delays, and lack intelligent assurance mechanisms.
Solution
TencentOS Server AI Enhanced Edition provides three core capabilities:
AI Ecosystem Adaptation Support
Broad Hardware Compatibility: As of July 2026, more than 40 verified GPU and NPU models are supported, including NVIDIA (A100, H800, and others) and AMD (300 series). For specific models and versions, refer to the compatibility list.
Mainstream Framework Integration: supports mainstream inference frameworks, training frameworks, and application frameworks. Agent frameworks can be deployed by following the documentation. For specific supported versions and prerequisite dependencies, refer to the compatibility list.
Rapid Deployment: The image comes preinstalled with commonly used AI ecosystem software and Codebuddy CLI. According to Tencent's internal compatibility validation and product capability list as of July 2026, it supports the installation and use of 1000+ agents and related components, including those preinstalled in the image, tested as installable, or provided with documentation-based deployment guidance. For details, refer to the compatibility list.
Intelligent Ops with Natural Language
AI Ops Diagnostic Capabilities: integrates 24 diagnostic capabilities and covers 9 major Ops domains, including disk and storage, performance analysis, memory diagnostics, system management, security compliance, fault recovery, and network diagnostics.
Performance Analysis and Optimization: provides CPU flame graphs, system call hotspot analysis, and process scheduling latency tracing to accurately identify performance bottlenecks such as high interface latency and slow containers.
Intelligent Resource Management: supports the diagnosis of issues such as OOM and memory leaks, helping improve the stability of long-running inference tasks.
On-Demand Loading Mechanism: AI matches diagnostic modules based on users' natural language descriptions, helping improve problem localization efficiency and interaction experience.
Kernel Crash Analysis Expert (AiCrasher)
Automated Analysis Process: By combining AI with the Linux crash tool, automated analysis can be performed for supported fault scenarios across 7 stages, from information collection and crash type identification to root cause localization, following a preset process.
Broad Crash Coverage: covers common kernel crash and exception scenarios such as HangTask, Soft/Hard Lockup, BUG_ON, NULL pointer, OOM, and SysRq-triggered events. For the specific supported scope, refer to the product documentation.
Community Patch Correlation: automatically searches for fix commits in the upstream Linux kernel community and checks whether the current distribution has been patched.
Structured Report Output: automatically generates analysis reports that include a complete evidence chain and remediation suggestions.
Benefit
|
Hardware Coverage | Supports 40+ mainstream GPU/NPU hardware models (as of July 2026, data sourced from Tencent internal compatibility testing). |
Ops Capabilities | 24 diagnostic capabilities cover 9 major domains (as of July 2026, data sourced from Tencent internal product capability list). |
Crash Analysis | Automatic analysis across 7 stages, covering multiple common kernel crash scenarios. |
Deployment Efficiency | Supports common AI environment configuration through preinstalled images and deployment documentation. Users still need to complete installation and verification based on hardware models, image versions, account permissions, and application dependencies. |
Typical Scenarios
|
AI model deployment | Supports deployment with the model, hardware, and version combinations listed in the compatibility list. Specific inference and fine-tuning capabilities are subject to the corresponding product documentation. |
Intelligent customer service system | Build an enterprise-grade intelligent Q&A and customer service system based on large models. |
Code-assisted development | Integrate Codebuddy CLI to improve development efficiency. |
AI Agent application | Supports deployment of the Agent framework with the frameworks and versions listed in the compatibility list. Specific details are subject to the corresponding product documentation. |
Supported Hardware and Framework Ecosystem
|
NVIDIA GPU | A100, H800, A800, L40S, and others |
AMD GPU | Details are subject to the compatibility list. |
Inference framework | vLLM, TensorRT-LLM, TACO-LLM, and others |
Training framework | PyTorch, DeepSpeed, Megatron-LM, and others |
Agent framework | OpenClaw, Hermes, and others |
TencentOS Server AI Enhanced Edition provides enterprises with adaptation, deployment, and Ops support for AI model development and deployment through its AI-oriented software and hardware adaptation, model deployment, and intelligent Ops capabilities, making it suitable for AI deployment scenarios.
Note:
Third-party products, models, frameworks, software, and trademarks mentioned in this document, including but not limited to DeepSeek, LLaMA, OpenClaw, Hermes, NVIDIA, PyTorch, and TensorRT-LLM, belong to their respective owners and are referenced only to illustrate the adaptation or compatibility scope of TencentOS Server. The specific supported versions, configuration requirements, and usage restrictions are subject to the compatibility list and product documentation. The specific meaning of "support" is as follows:
"Verified Compatible": Compatibility verification has been completed in Tencent's internal test environment.
"Deployable via Documentation": Users need to install and configure it by referring to the official documentation.
"Preinstalled": The software is included in the image by default.
Statistics such as the number of hardware adaptations, diagnostic capabilities, and documents are current as of July 2026. These figures may change in future versions, so always refer to the latest product documentation.
Summary
TencentOS Server provides relevant OS capabilities for core scenarios such as system migration, database foundations, cloud native support, and LLM inference acceleration, and has accumulated large-scale operational practices. Whether it is handling migrations due to the discontinuation of legacy systems, improving database and cloud platform performance, optimizing cloud native resource efficiency, or accelerating private deployment of large models, TencentOS Server delivers OS-level support for enterprise digital transformation with solid technical capabilities. For quantitative data mentioned in this document, see the test statements in the descriptions of each scenario.