Interpretation of the "OS Industry Report": The Technological Breakthrough and Ecological Leap of Domestic Operating Systems

Under the dual wave of digital transformation and domestic substitution, server operating systems are transitioning from "usable" to "user-friendly," becoming the core foundation of the industry. At the recent 2025 Longxin Operating System Conference, the China Intelligent Computing Industry Alliance released the "Domestic Server Operating System Development Report (2025)," which predicts that in the next three years, the domestic OS localization rate will increase from 35% to 65%, and the market scale will exceed 30 billion yuan. It clearly identifies key trends in OS evolution such as "cloud + AI," "open-source ecosystem," and "RISC-V architecture." The seventh issue of "AI Evolution" invited three industry experts: Zhang Pengcheng, Product Director of Alibaba Cloud's Basic Software Department; Xia Zhenchun, Deputy General Manager of ZTE New Benchmark Company; and Cui Baiqie, Chief Analyst of InfoQ Research Center, to deeply analyze the core data and insights from the report, rigorously dissecting the industrial growth logic, technological breakthrough paths, ecological challenges, and future development blueprint of domestic OS, decoding the key to the transition of domestic operating systems "from catching up to leading." https://static001.geekbang.org/wechat/images/33/3332cd0e1876d9b1ef16ab4a13dda31d.png The following is the edited transcript of the expert interview. Core Insights from the Report — Current Status and Data Interpretation of Domestic OS Ecosystem Q1: What are the sources and research methods for the data in the report, such as Longxin operating system's installed base exceeding 10 million, market share nearing 50%, and the domestic OS localization rate doubling with the market scale exceeding 30 billion? What are the core factors supporting these performances? @Cui Baiqie: The data in the report does not come from a single source; it is primarily based on publicly available industry data, data released by manufacturers and the Longxin community, and is supplemented by systematically designed survey questions and samples to ensure data validity. There are three core factors supporting the rapid growth of the localization rate: first, after CentOS ceased operations, original users needed to seek alternative solutions, and domestic manufacturers quickly followed up with relevant solutions; second, the localization demand in industries related to finance, energy, and government enterprises is relatively rigid; third, "cloud + AI" has opened up new battlegrounds, providing new development space for domestic operating systems. The high market performance of the Longxin operating system is partly due to the scale advantage of exceeding 10 million installations by 2025; on the other hand, survey results show its outstanding attractiveness in current usage and future migration willingness, with users rating its security and underlying native support highly, and its "cloud + AI" development direction has also enhanced market confidence. Q2: The report indicates that users generally recognize the stability, security, and performance optimization capabilities of domestic OS. Is this consistent with the actual feedback from ZTE's industry implementations? How do these changing demands drive ZTE's technological iterations? ** @Xia Zhenchun: ZTE has been deeply engaged in the telecommunications industry for many years. It started the research and development of its own operating system kernel in 2002, and it has been 23 years since then. The new pivot server operating system has been stably running in key areas such as telecommunications, power, government affairs, and finance for over 20 years. By early 2025, the number of releases is expected to reach around 100,000 sets. These data and cases fully confirm users' recognition and trust in domestic operating systems, which is highly consistent with the report's conclusions. The stringent demands of industry users are the core driving force for technological iteration: first, the requirements for uninterrupted operation 24/7 under extreme loads, fault switching in seconds, and kernel security hardening drive us to continuously optimize core capabilities; second, the demand for low latency and high concurrency in the telecommunications field prompts ZTE to collaborate with communities like Longxia to optimize kernel scheduling and network protocol stacks, enhancing system throughput; third, the rigid demand for data sovereignty and security drives us to strengthen compatibility with domestic chips, support for confidential computing, and supply chain security certification, achieving the evolution of systems from usable to trustworthy, controllable, and maintainable enterprise-level platforms, forcing technological innovation through industry pain points. Q3: Why can open source become the core support for the large-scale development of domestic OS? What aspects still need to be broken through in the domestic open source community compared to the international open source ecosystem? @Cui Baijie: In the current basic software field, it is difficult to rely solely on a single vendor to promote kernel evolution, implement solutions, and aggregate ecosystems. The open source ecosystem can unite the forces of operating system developers, hardware vendors, and downstream users to form a development synergy, which is the key reason the report lists open source as a core support. Compared to international open source ecosystems like Apache and Linux, the domestic open source community needs to break through in two aspects: first, to form more technical consensus, including the unification of technical standards, interface standards, technology stack standards, and evaluation standards, allowing industry forces to focus; second, to enrich contribution methods, encouraging developers to actively participate in or even lead rule-making, feeding back the collaborative development of the operating system and basic software ecosystem. @Xia Zhenchun: Open source aligns with the domestic basic software industry's reality of "late start, little accumulation, and fragmentation," and can solve the problems of repeated R&D and high costs for enterprises. By gathering the wisdom and achievements of global developers, domestic operating systems can quickly achieve a technological leap from 0 to 1, reducing R&D costs for enterprises and individuals. Most domestic open source communities adopt a "community co-construction + commercial distribution" dual-drive model to accelerate the compatibility adaptation of software and hardware ecosystems and the formulation of industry standards, but there is still room for improvement: first, insufficient governance discourse power, often following positions in core project decisions and technical directions; second, ecological collaboration is fragmented. Although major communities are actively promoting kernel unification, inconsistent interface and compatibility certification standards among vendors affect user experience; third, there is a certain lag in the construction of basic capabilities. Security audits, vulnerability responses, long-term support, and incentive systems have not yet formed standardized systems in line with international practices, and the RISC-V architecture ecosystem incubation is still in its early stages, needing to achieve systematic upgrades from code contributions to standard formulation and ecological operation Q4: As the initiator of the Longxi community, how does the "open source + commercial" collaborative model land in the community? What support can this model provide for the upcoming market scale of 30 billion? @Daniel Zhang: The development of Alibaba Cloud in the operating system field is a reflection of "originating from open source, contributing to open source, and benefiting from open source." From the early demand of Taobao to "get rid of IOE," which spurred kernel technology accumulation, to the service upgrades brought by cloud business expansion, and to new opportunities in AI scenarios, Alibaba has always been sedimenting and growing within the open source ecosystem. From the community perspective, the "open source + commercial" collaborative model lands through two major paths: first, at the community governance level, the Longxi community has established a common governance mechanism based on core council units over the past five years, gathering enterprises, developers, and industry groups through operations like the Longteng Plan, Security Alliance, System Operation and Maintenance Alliance, and Intelligent Computing Alliance to build an open source collaboration model; second, at the commercial ecosystem level, forming a pattern of "upstream community version + downstream derivative commercial distribution version," where vendors like ZTE and Alibaba Cloud serve specific customer groups through commercial distribution versions, gaining commercial benefits in services and specific capabilities, achieving a virtuous cycle between the open source ecosystem and commercial operations. This model can provide core support for the 30 billion market scale: on one hand, open source lowers technical thresholds and R&D costs, expanding market participation; on the other hand, commercial distribution versions meet the personalized needs of different industries, realizing value monetization, and the synergy of both drives continuous market scale growth, which is consistent with the development paradigm of the international open source industry. Trend Implementation Practice - Observing Community/Enterprise Actions from Report Trends Q5: The report points out that "cloud + AI" is the primary innovation direction for domestic OS. Why? What explorations has Alibaba Cloud undertaken? @Daniel Zhang: In the wave of AI, the construction of AI infrastructure is accelerating, and the application of AI intelligent entities is flourishing, driving a surge in computing power demand and technological stack innovation, bringing significant opportunities for the integration of "cloud + AI" and operating systems. For example, cloud platforms that typically operate at scale can create tens of millions of virtual machines cumulatively, with some AI startups creating over 80 million virtual machines in a single year. This high-frequency, elastic demand for computing power promotes the deep integration evolution of operating systems towards "cloud + AI." Against this backdrop, Alibaba Cloud, in collaboration with the Longxi community, proposed a "dual-loop" technical route of System for AI and AI for System: System for AI: Combining AI infrastructure, generative AI-related training inference frameworks, runtime sandboxes, etc., with the operating system's resource optimization, compilation optimization, operator optimization, and framework optimization to enhance the startup speed and concurrent elasticity of AI applications, reducing the unit cost of model training and inference; AI for System: Applying models to the operation of the operating system itself, including dynamic system tuning in different scenarios through algorithms, incubating OS copilot and AI intelligent entities to simplify operation and maintenance processes, and integrating intelligent entity capabilities to enhance vulnerability response and patch repair efficiency, making the operating system safer and more stable This integration is not a simple functional overlay, but rather a deep interaction between AI middleware capabilities and the operating system kernel and surrounding modules, forming a synergistic effect. Q6: The report points out the pain points of domestic OS, such as "difficulty in adapting to AI scenarios and lack of industry-wide unified standards." Can you discuss the problem-solving path in conjunction with ZTE's new pivot OS V7's AI adaptation practices? @Xia Zhenchun: The new pivot server operating system V7, to be released in 2025, has fully adapted to mainstream AI hardware (CPU, GPU, NPU, DPU, etc.), large models, and mainstream frameworks both domestically and internationally, significantly lowering the deployment threshold for AI applications on domestic operating systems and effectively solving the fragmentation issue of heterogeneous hardware and model adaptation. Additionally, the telecom scenario TFCE evaluation dataset that we open-sourced to the community serves as the first public evaluation benchmark for AI operations and maintenance in the telecom core network, filling the gap in systematic capability assessment in real business scenarios within the industry, providing a reusable and quantifiable unified standard for the tuning validation and comparative analysis of AI models, servers, and operating systems. These practices have also achieved a feedback loop to the community ecosystem: first, the dataset can be adopted by the community as a reference benchmark for AI infrastructure evaluation, enhancing the AI performance evaluation toolchain level of the Longxin distribution; second, ZTE's accumulated AI tuning experience in telecom scenarios will be transformed into general kernel patches; third, by establishing SIG groups and sharing results such as confidential computing container deployment solutions, we assist the Longxin community in building an "industry scenario + open-source collaboration" ecological closed loop. Q7: What historic opportunities has the RISC-V architecture brought to domestic OS? What technical challenges have ZTE's related practices solved? @Xia Zhenchun: The RISC-V architecture is evolving from embedded systems to high-performance computing, with an explosive demand for server-level RISC-V, and the potential of a collaborative open-source ecosystem for software and hardware is enormous, providing historic opportunities for domestic OS to break through traditional technical paths, and ZTE's practices fully align with this trend. In 2025, ZTE will deeply participate in the Longxin community RISC-V SIG 2.0 planning, promoting the landing of the Anolis 23 RISC-V preview version, focusing on solving three major technical challenges: first, kernel and architecture adaptation, completing deep optimization of the kernel for RISC-V server-level instructions (vector extensions, virtualization extensions), filling the gap in kernel support for high-performance RISC-V servers in domestic operating systems; second, building a collaborative ecosystem of software and hardware, working with chip manufacturers to streamline the compilation toolchain, driver framework, and system scheduling mechanisms, achieving pre-research-level compatibility for RISC-V server SoCs with multiple systems and unscaled applications, accelerating the hardware definition, system adaptation, and ecological verification cycle; third, collaborative open-source governance, promoting cross-community technical consensus through RISC-V SIG 2.0, unifying key interface standards for RISC-V server scenarios, alleviating ecological fragmentation issues, and laying the foundation for building a self-controllable, open, and collaborative RISC-V server operating system ecosystem Pain Point Tackling - Core Challenges and Solutions for Domestic OS Q8: What are the specific manifestations of the two major pain points, "ecological fragmentation and core technology shortcomings," in enterprise implementation? Which industries are most affected? @Cui Baijie: The specific manifestations of ecological fragmentation are very prominent: first, multiple operating systems coexist within enterprises, due to strategic choices to avoid binding with a single vendor, the continuation of old and new business systems, and compatibility needs for heterogeneous chip architectures; second, the compatibility challenges brought by heterogeneous chip architectures, with the emergence of various architectures such as TPU, DPU, and FPGA in AI scenarios, which place higher demands on operating system compatibility; third, the difficulty of software-hardware collaboration, where enterprises focus most on the compatibility of the operating system with domestic databases, middleware, and mainstream open-source applications, even surpassing traditional core demands such as security updates and stability, becoming the core challenge of ecological collaboration. Key industries such as finance, energy, and telecommunications are most affected. These industries have complex operations, large system inventories, and diverse heterogeneous resources, requiring high compatibility, stability, and security from operating systems. Ecological fragmentation and core technology shortcomings will directly impact business continuity and the process of domestic substitution. Q9: How does the Longxin community's collaborative model alleviate the problem of ecological fragmentation? What is the division of labor and cooperation between Alibaba Cloud and ZTE in this? @Zhang Pengcheng: The Longxin community alleviates ecological fragmentation through a unified baseline and collaborative adaptation model. Currently, thousands of partners are involved in co-construction, with two core practices: first, establishing an upstream driving baseline, allowing hardware vendors to achieve compatibility with all operating system distributions within the Longxin ecosystem through a single adaptation, reducing repeated adaptation costs and enhancing ecological convenience; second, creating a "one test, multiple certifications" model, using the Longxin community version as the upstream, standardizing specifications, allowing adapted software and hardware to share adaptation effects across multiple distributions, and finding the greatest common divisor for ecological compatibility. @Xia Zhenchun: Collaboration is the core of ecological development. The Longxin community unites five major mainstream communities to standardize the kernel version baseline and interface standards, addressing the fragmentation of multiple system versions and upstream-downstream compatibility issues at the foundational level, providing customers with a unified technical base. Alibaba Cloud and ZTE have distinct focuses in their division of labor: ZTE, as a key industry practitioner, deeply engages in high-security scenarios such as telecommunications and power, adapting the Longxin kernel to critical devices and systems like 5G base stations, gateways, power forecasting platforms, and grid systems, feeding back issues discovered during implementation and accumulated experiences to the community, enriching the community's industry application scenarios; Alibaba Cloud plays a core role in community governance, technical base construction, and ecological aggregation, forming a paradigm of "unified foundation, diverse upper layers, and software-hardware collaboration," promoting the transition of domestic operating systems from breakthrough points to ecological common prosperity. Q10: What core technologies has ZTE output in the Longxin Security Alliance? How does it balance security protection and system performance in key industry implementations? @Xia Zhenchun: As a core participant in the Longxin community security alliance, ZTE is deeply involved in vulnerability discovery and management. It has achieved practical results in areas such as hardware trusted execution environments, confidential computing frameworks, intrinsic security architecture design, and software supply chain security lifecycle management. It plans to output public welfare and common elements to the security alliance in the future The new pivot server operating system has achieved end-to-end security hardening from code source review, vulnerability closed-loop management to runtime protection. It balances security and performance in key industries through three major strategies: first, soft and hard collaborative optimization, deeply integrating the RISC characteristics of domestic chips with the operating system kernel, combined with lightweight container isolation technology, ensuring full-link encryption of data transmission, storage, and computation while controlling performance overhead within 5%; second, intelligent resource scheduling, dynamically allocating resources through algorithms to avoid increased latency caused by encrypted computing; third, rapid fault switching, achieving automatic switching within 30 seconds in 5G core network signaling processing while ensuring that user privacy data does not leave the domain, truly realizing "security without performance degradation, performance without compromise," responding to the report's requirements for enhancing confidential computing compatibility and strengthening endogenous security. Q11: What is the biggest technical challenge faced by Alibaba Cloud in the development of the Longxin community kernel? How does the community collaborative attack mechanism improve R&D efficiency? @Zhang Pengcheng: In the development of the Longxin kernel, Alibaba Cloud focuses on four dimensions: "stability, security, performance, and cost," making it difficult to define a single "biggest challenge," as each dimension has targeted technical breakthroughs: In terms of stability, to support the parallel operation of millions of servers and handle hundreds of billions of application loads, we collaborate with chip manufacturers like Intel and Haiguang to enhance system reliability and problem diagnosis capabilities using Rust features; In terms of security, we focus on the research and development of confidential computing technology, forming joint solutions with partners such as the Coco community under CNCF to ensure data security and privacy protection for cloud customers; In terms of performance, we collaborate with chip manufacturers to optimize specific scenarios such as web applications and databases, utilizing kernel-level SMC capabilities to achieve RDMA performance effects under TCP/IP networks, enhancing the cost-effectiveness of customer computing power; In terms of cost, we support the mixed deployment of online and offline tasks through kernel resource isolation and QoS assurance technologies, achieving fine-grained scheduling and balanced resource usage. Of course, the community collaborative attack mechanism has also significantly improved R&D efficiency: on one hand, issues in large-scale application scenarios can be quickly fed back to the community, clarifying R&D directions; on the other hand, the participation of multiple parties such as chip manufacturers and application vendors in joint research avoids a single enterprise working in isolation, accelerating the resolution of core technical shortcomings and making R&D results more aligned with actual needs. Future Outlook - From Segmentation to Generalized Industrial Evolution Q12: What key nodes must the domestic OS break through to move from niche areas to the general market? What role should the open-source community play? @Cui Bai Jie: The current operating system market has entered an ecological competition stage, and the key nodes for moving from niche to general markets lie in ecological aggregation and value realization: first, gathering diverse ecological partners, covering participants across the entire industry chain, including hardware, software, and services; second, reducing user selection and migration costs to form a sustainable ecological system; third, deeply binding technology with scenarios to provide practical solutions for different industries, enhancing the actual value of the ecosystem @Daniel Zhang: From the segmented to the general market, it is necessary to balance the evolution of the existing market and the expansion of the incremental market: In terms of the existing market, the core challenge is to achieve smooth replacement by enhancing ecological compatibility and optimizing migration tools, thereby reducing the risk for enterprises to replace their existing systems (such as CentOS), as changes are often the main inducement for IT system instability; In terms of the incremental market, cloud and AI scenarios are core opportunities. These scenarios have no historical burdens and can quickly adapt to new operational system requirements, and the overseas market has already been validated (such as the widespread application of Ubuntu in AI scenarios). New scenarios such as domestic GPUs, large models, and AI intelligences will drive operating systems to naturally gain market share. The open-source community should play the role of a "technical bridge": first, to assist in the smooth migration of the existing market by lowering replacement barriers through unified standards and optimized tools; second, to accelerate the implementation of technologies in the incremental market by validating new scenario technical solutions through community collaboration, promoting the integration of cutting-edge technologies such as cloud, AI, and RISC-V with operating systems. @Xia Zhenchun: The industry needs to break through two key nodes: first, a systematic breakthrough in core technologies, shifting from single scenario optimization to multi-scenario balance, forming generalized solutions in kernel-level performance, concurrency capabilities, and distributed architecture to adapt to the needs of different industries; second, the integration of the industrial ecosystem to solve issues of heterogeneous resource compatibility, unified scheduling, and hardware-software collaborative adaptation, breaking the fragmentation problem. The open-source community represented by Dragonfly should undertake two core roles: first, as a technology aggregator, solving technological fragmentation by unifying kernel baselines, cross-community technical collaboration, and incubating standardized distributions; second, as an ecological collaboration hub, gathering wisdom from all parties based on an open cooperation model, accelerating hardware-software adaptation, and promoting the implementation of cutting-edge technologies such as AI, edge computing, and RISC-V. Q13: The report predicts that Dragonfly's market share in the "cloud + AI" incremental market will exceed 50%. What is the current progress? Is this prediction achievable? @Daniel Zhang: This prediction is based on industry trends and actual data support, achieving a high degree of certainty: From the perspective of industry trends, the accelerated development of cloud and AI has become a foregone conclusion. Four to five years ago, a definitive turning point appeared in the cloud market, with the proportion of new servers aimed at the internet and cloud continuously increasing. The distributed and cloud-native characteristics of AI scenarios are deeply bound to operating systems, forming a trend of computing power evolution dominated by "cloud + AI"; From the perspective of actual progress, Alibaba Cloud, as a leader in the cloud market, has become the operating system with the highest retention on the cloud, and in the past year, the proportion of new computing power on the cloud choosing Alibaba Cloud's Dragonfly version of the operating system has reached 80%-90%; From the perspective of future potential, with the continuous explosion of domestic intelligent computing architectures, large models, and AI intelligences, Dragonfly's technological adaptation and ecological advantages in the "cloud + AI" field will further expand. A market share exceeding 50% is a natural result of industrial development and will gradually materialize with the explosive speed of the incremental market Q14: What suggestions do you have for the ecological development of domestic open-source OS and for enterprises or developers who want to join the community? @Xia Zhenchun: The core suggestion is "focus on technological breakthroughs + strengthen ecological integration." Technically, it is necessary to achieve systematic improvement of core capabilities in multiple scenarios, and ecologically, it is essential to solve the fragmentation problem and form a collaborative force. There are three suggestions for enterprises and developers: First, start with contributions; whether it is submitting driving patches, optimizing documentation, or fixing CVE vulnerabilities, these are effective starting points for participating in ecological construction. Second, embrace openness while adhering to security; when using open-source components, it is necessary to establish code review, vulnerability response, and supply chain audit mechanisms to ensure compliance and trustworthiness. Third, focus on scenario collaborative innovation; combine industry pain points with community co-construction of SIG groups, allowing open-source results to truly solve industrial problems and achieve a leap from participant to leader. @Zhang Pengcheng: The core is "root down + grow up": rooting down means building robust standards and unified specifications, uniting domestic technical forces, and solidifying the technological foundation for independent innovation; growing up means relying on scenario innovation to drive technological evolution, providing momentum for technological innovation through killer applications and attractive scenarios, forming a virtuous cycle of ecological development. We hope that enterprises and developers will use the community as a carrier to deeply participate in technological co-construction and scenario implementation, contributing at the technical level and exploring value at the scenario level, jointly promoting the stable and long-term development of the domestic open-source OS ecosystem. Q15: What industrial-level collaborative efforts are needed to achieve the goal of a 65% localization rate? @Cui Baijie: Achieving this goal is a systematic project that requires collaboration across policies, technology, and ecology: On the technical level, it is necessary to continuously optimize the compatibility and stability of the existing market while keeping up with emerging trends such as "cloud + AI" to achieve iterative upgrades of core technologies; On the ecological level, it is essential to further aggregate industrial forces, reduce migration costs, better adapt technology to actual scenarios, and enhance user experience and recognition; On the policy level, the continuous promotion of policies related to Xinchuang will provide significant support, and it is expected that by 2027, more Xinchuang results will be realized, providing policy guarantees for the increase in localization rates. Conclusion: The rise of domestic OS begins with collaboration and is achieved through diligence The blueprint depicted in the "Domestic Server Operating System Development Report (2025)" of "65% localization rate" and "30 billion market scale" is not an unattainable goal but rather an inevitable result of the collaborative efforts of the open-source ecosystem, breakthroughs in "cloud + AI" technology, innovations in RISC-V architecture, and practical implementations by enterprises. Currently, domestic operating systems have gained market recognition in core capabilities such as stability, security, and compatibility, achieving large-scale implementation in key industries, while ecological fragmentation and core technological shortcomings are gradually being resolved through community collaboration. From segmentation to generalization, from usable to user-friendly, the evolution of domestic operating systems requires continuous technical breakthroughs, open collaboration at the ecological level, and guiding support at the policy level. In the future, with the continued efforts of open-source communities like Dragonfly, deep participation from companies like Alibaba Cloud and ZTE, and collaborative efforts from industry, academia, and research, domestic operating systems will become the core foundation supporting the digital economy and localization replacement And occupy an important position in the global infrastructure software market. Click to view the complete live replay video～ Click to get the download link!