In the post-Moore era, the primary driver of computing power growth has shifted from semiconductor process miniaturization to full-stack software-hardware co-optimization. During the domestic substitution of computing power in China, technical bottlenecks exist, including the absence of full-stack optimization toolchains, insufficient technical accumulation of key system software, and high costs for the migration of legacy applications. This paper proposes a method of data-driven system optimization, which strives to transform performance optimization from empirical attempts into a quantitative and reproducible engineering approach, simultaneously emphasizing the system perspective of vertical full-stack and horizontal scaling. Based on the proposed method, this paper introduces the research progress in microarchitecture measurement, compiler optimization and binary translation, and outlines future technological development trends. This work provides valuable references for promoting the independent and controllable development of China’s domestic computing power ecosystem.
As database systems continue to evolve toward distributed architectures, cloud-native deployments, hybrid transactional/analytical processing workloads, multimodal data management, and agent-driven access, their evaluation frameworks—including target objects, workload characteristics, and quality metrics—have changed significantly. Conventional evaluation methodologies are primarily designed for relational databases with fixed query patterns, stable workloads, and single-dimensional performance metrics, thus rendering them insufficient for comprehensively characterizing modern database systems operating under complex functionalities, hybrid workloads, abnormal operating conditions, ecosystem migration, and security constraints. Hence, this paper presents a systematic taxonomy of database-evaluation techniques across five key quality dimensions: functional correctness, performance, availability, compatibility, and security. Furthermore, the research progress, application scenarios, and limitations of representative evaluation approaches are reviewed. Subsequently, challenges are introduced by large language models and agent-driven database interactions. We argue that future database evaluation must shift from isolated functionality testing to end-to-end quality validation, and from static, benchmark-based assessment to dynamic scenario simulation. Furthermore, evaluation frameworks must encompass emerging evaluation targets, such as multimodal queries, agent interfaces, structured result generation, and end-to-end security verification. This survey aims to provide a comprehensive reference for academic researchers and engineering practitioners in modern database evaluation.
Blockchain is a key technology for building trust infrastructure based on data and mathematics. With ever-deeper applications in economy, livelihood, and other fields, the number of users has expanded rapidly, and task logic has become increasingly complex. How to enhance the scalability of blockchain systems thus becomes critical. Traditional distributed database systems only need to handle fail-stop faults, whereas blockchain systems must tolerate Byzantine faults, which results in significantly lower execution performance. System architecture is critical for blockchain performance, and architectural innovation is particularly important for scalability. This paper reviews the impact of architectural evolution on blockchain scalability, conducts an in-depth analysis of trust relationships among nodes, and proposes a hierarchical hybrid sharding architecture that organizes and manages shards in a layered manner, significantly enhancing system scalability. Further improvements in blockchain scalability include trust mechanism design, modular design, hardware and software co-design, on-chain and off-chain coordination, and deep integration with artificial intelligence.
As large language model (LLM) equips agents with stronger semantic understanding and task execution capabilities, multi-agent systems are transitioning from closed intranets to open networks. They show great potential for cross-institutional financial auditing, distributed scientific computing, decentralized intelligent services, and related scenarios. However, open environments often lack mutual trust, involve heterogeneous interests, and operate without a central authority. As a result, existing systems struggle to support verifiable collaboration, reproducible and recoverable failure handling, and accountable agent behavior. The objective of this study was to systematically review the key challenges of multi-agent collaboration in open environments. These challenges include the tension between the semantic correctness of state transitions and the verifiability of logs, trusted storage and version consistency of long-term memory, and contribution attribution and incentive compatibility among self-interested agents. Based on this analysis, a theoretical framework grounded in blockchain and verifiable computation is proposed. This framework includes an integrated on-/off-chain architecture, verifiable fault-tolerant state machines, verifiable storage and consistency maintenance for semantic memory, manipulation-resistant attribution, and incentive mechanisms. This research will foster deep cross-fertilization among distributed systems, formal verification, game theory, and LLM uncertainty, providing fundamental theoretical support for trustworthy collaboration in the “AI+” initiative.
Cross-organizational data application programming interface (API) calls frequently involve sensitive identity and authorization information of individuals and legal entities. Account credentials, Internet Protocol (IP) allowlists, and ordinary access tokens mainly authenticate callers, but cannot prove that each call has obtained valid authorization from the data subject, constrain critical business parameters, or support end-to-end traceability. To address these problems, this paper proposes a trusted access mechanism coordinated by a trusted access agent (TAA), a trusted access platform (TAP), and a trusted access gateway (TAG). Referring to attribute-based access control (ABAC) and the extensible access control markup language (XACML), the mechanism models trusted data access as six-dimensional constraint matching across subjects, applications, terminals, resources, purposes, and validity periods. A trusted access token, which can be represented using a format such as JSON web token (JWT), binds authorization evidence to access claims, while the resource-side TAG performs signature verification, validity checking, scope matching, replay prevention, and audit logging. Taking the socialized application of electronic certificates as a typical scenario, this paper analyzes component mapping, cross-network deployment paths, and typical abnormal access conditions. The analytical results demonstrate that the mechanism strengthens subject-authorization binding, request-consistency verification, least-privilege control, and end-to-end traceability for protected data API calls.
中文核心期刊
