Capitalonline Data Service Co., Ltd.
首都在线
300846
Shenzhen Stock Exchange
Company Profile
i. Solid customer base and close customer relationships After years of diligent work, the Company has accumulated a large and widely distributed customer base. With long-term efforts and accumulation, we have built a solid foundation of customers and established long-term partnerships with many leading enterprises in various industries. The broad customer coverage enables the Company to gain in-depth insights into the diverse needs of customers across different industries, providing a rich practical basis for the optimization and upgrading of products and services. ii. Rich product matrix and scenario-based solutions The Company offers a variety of cloud computing products and integrated solutions, allowing customers to freely choose the most suitable combination of cloud products and solutions. We have fully developed core cloud computing products in-house and, through ecosystem partnerships, further enriched our product system, forming a product matrix that meets the full lifecycle needs of customers. Focusing on industries such as large models and AIGC applications, government affairs, finance, education, media, and the internet, the Company has launched a series of solutions that can be rapidly deployed in combination with customers' actual business scenarios. iii. Core technological accumulation Since its inception, the Company has attached great importance to independent research and development of core technologies, having accumulated core technological assets such as the Global Private Network and the Cloud OS. The Global Private Network enables high-speed, secure global network interconnection. The Cloud OS meets customers' core needs for high-performance computing and elastic business services. The Global Private Network is a dedicated line solution implemented by the Company based on a globally multi-ring redundant network. Any virtual data center deployed globally can be connected to the cloud interconnection network service, enabling high-speed, stable, and secure internal network interconnection within the virtual data center. Currently, the Company’s global cloud interconnection network covers major countries and regions around the world, including Mainland China, Hong Kong, Japan, Singapore, the United States, and Germany, forming high-speed and secure global network interconnectivity. The Cloud OS is an in-house developed cloud computing service platform that offers a complete set of cloud resource management capabilities, including virtualization, distributed storage, SND network, database services, big data services, and management control consoles, characterized by security, stability, simplicity, and efficiency. Users can quickly and securely build and deploy business applications through the management console or APIs, as well as set up personalized PaaS and SaaS services on the platform. The Automated Operations and Maintenance Platform is an important support to ensure the stable operation of the cloud platform. The Company's self-developed automated operations and maintenance platform can provide multi-dimensional real-time monitoring of various resources within the cloud platform, such as servers and network devices. By setting up various monitoring indicators, it can promptly capture performance fluctuations of resources and address potential faults. Meanwhile, based on big data analysis and artificial intelligence algorithms, the platform integrates various data sources, including log information and performance indicators, and can quickly locate fault sources, avoiding the complexities and inefficiencies of traditional manual troubleshooting. iv. Continuous R&D investment and technological innovation capabilities The Company has always placed great importance on R&D investment, with a total R&D expenditure of RMB218.98 million over the past three years, forming a technological accumulation with certain advantages. We have a strong professional R&D team, with R&D/technical personnel accounting for over 50%, and the majority of core R&D staff having more than 10 years of R&D experience. All core R&D staff hold degrees from China’s Project 211 universities or equivalent top-tier institutions. By gathering a group of high-level researchers, the Company continuously improves its independent R&D and innovation capabilities. v. Excellent customer support and service capabilities The Company has developed a complete customer service system, providing users with "7×24" hour online technical support services. VIP customers are offered dedicated service channels, one-on-one exclusive service managers, service engineers, and on-site engineers, among other services. This has gained widespread recognition from customers and established a good reputation in the industry.
Full description
The Company is principally engaged in providing cloud services, communication network services, IDC services, and integrated solutions, encompassing both general computing and intelligent computing. We are committed to constructing stable and efficient cloud-network integrated services for industries such as large models and AIGC applications, government affairs, finance, education, media, and the internet, meeting the service demands for computing power and network capabilities during digital transformation in different industries. 1. Intelligent computing cloud business (1) Intelligent computing container cloud product The intelligent computing cloud container product adopts a deeply restructured cloud-native technology stack. Based on the Kubernetes Operator framework and CRD extension mechanism, it builds a fully managed container service cluster that realizes Serverless lifecycle management for container instances. The product kernel integrates Containerd (a high-performance container runtime) and eBPF-based network acceleration technology. Through coordination between the hardware virtualization layer and container orchestration layer, a high-density computing platform with microsecond-level scheduling delay is constructed. This platform can support distributed AI training tasks on multi-thousand-GPU clusters. (2) Vector database product The vector database is built on a high-dimensional data computing engine with a deeply optimized architecture. It builds a fully distributed storage layer based on NVMe over Fabrics and RDMA network acceleration technology, supporting sub-millisecond nearest neighbor retrieval for PB-level vector data. The core engine integrates the hierarchical navigable small world (HNSW) index and product quantization (PQ) algorithm, combined with an adaptive pruning strategy, achieving a hundredfold compression ratio under a recall rate of over 98%, and improving query throughput by 3-5 times compared to traditional ANN algorithms. The system incorporates a hybrid computing-storage integrated design, optimizing GPU-accelerated inference engines and CPU SIMD instruction sets, and supports FPGA card-based vector pre-filtering, enabling concurrent processing with a QPS of over 500,000 queries per node on a 4096-dimensional vector dataset. Through vector computation pushdown and native support for the Apache Arrow memory format, zero-copy data interaction is achieved with AI frameworks such as PyTorch and TensorFlow, ensuring that end-to-end retrieval latency is controlled within 1.2ms. This architecture adopts a multi-tenant security isolation scheme and realizes vector encrypted retrieval via the SGX 2.0 trusted execution environment, complying with the IEEE 2953 vector computation standard. It supports Kubernetes dynamic scaling policies and can be linearly expanded to a thousand-node scale, with single-cluster throughput reaching hundreds of millions of queries per second. (3) MaaS platform The MaaS platform, through a unique computing power management and transaction mechanism, gathers multiple computing power resources, achieving efficient management and scheduling, and promoting the co-construction and sharing of computing power. The platform's core services integrate mainstream large model APIs from both domestic and international sources, including DeepSeek in China and Llama and Bloom abroad, providing users with one-stop model invocation, training, and fine-tuning capabilities. The MaaS platform supports multi-scenario adaptation, covering fields such as natural language processing, computer vision, and speech recognition, enabling enterprises to quickly deploy AI applications. (4) AI Infra platform The AI Infra platform primarily provides algorithm engineers with a one-stop model development and fine-tuning environment. It integrates high-performance computing resources, distributed training frameworks, automated tuning tools, and an open model library, significantly reducing the development threshold and improving research and development efficiency. Through cloud-native and containerized technologies, the AI Infra platform enables elastic resource scheduling and fast environment deployment, supporting multi-scenario adaptation and flexible deployment. In addition, it provides collaboration and sharing features and an open developer community, helping algorithm engineers efficiently complete the full process from model development to deployment, driving the rapid application and innovation of AI technology. 2. Computing cloud business Computing cloud products The Company provides cloud host and bare-metal products, with CPU as the primary computing resource, as well as network, storage, and other products and services. The self-developed GIC platform allows global cloud services and network resources to be activated, managed, and scheduled within seconds. The platform covers core regions globally, including North America, South America, Europe, Asia-Pacific, Mainland China, Hong Kong, Macau, and Taiwan, spanning over 50 countries and regions, allowing multi-point global deployment to be completed in just five minutes. The cloud host product is based on a deep customized KVM and a dual-engine virtualization cloud host architecture implemented on VMware enterprise platform, along with a high-performance distributed block storage collaborative architecture, providing a high-performance and highly compliant digital base for Chinese internet companies going abroad. The single-cluster availability is 99.995%, supporting scaling to thousands of CPU/GPU cores in seconds; block storage employs dual replicas + intelligent hot/cold layering, with data persistence reaching 99.9999999%, reducing storage costs by 50%. Through reinforcement learning scheduling algorithms, resource utilization exceeds 85%. The platform empowers industries such as cross-border e-commerce, audio/video, gaming, and advertising to expand globally, ensuring high-concurrency scenarios for hundreds of millions of users and providing a secure, elastic, and cost-effective full-stack cloud infrastructure for overseas operations. The platform also leverages its global network interconnection advantage to provide customers with low-latency, high-reliability internal network interconnection capabilities, helping them achieve global network architectures based on Layer 2 and Layer 3 networks. The Company’s global network is already interconnected with most global exchange centers and well-known internet service providers, reaching 80% of the global population within 20ms latency. 3. IDC business The IDC business mainly provides customers with server hosting, rental, and related value-added services such as network access and security protection, ensuring the security of enterprise data and the reliability of business systems.