Menglong Li

In cloud-network convergence environments, Virtual Network Embedding (VNE) confronts three fundamental challenges: resources heterogeneity, services dynamicity and policies diversity. Traditional exact algorithms, (meta)heuristic methods and learning-based approaches are respectively con strained by computational complexity, limited generalization capability and data dependency, rendering them inadequate for addressing these challenges effectively. Although Large Language Models (LLMs) demonstrate remarkable capabilities in semantic understanding and reasoning, their direct application to VNE still suffers from difficulties in cross-modal coordination, insufficient solution reliability and the inability to quantify business intents. To address these limitations, we propose FLEXVNE, an LLM-driven flexible VNE framework that incorporates three key innovations: (i) a coordinated feature encoder that fuses network graph structure with natural language descriptions to precisely characterize heterogeneous resources; (ii) a VNE head equipped with dynamic masking and hierarchical pol icy networks, embedding hard constraints directly into the decision process to ensure feasible and reliable embedding solutions; (iii) an intent-driven policy controller that transforms ambiguous natural language requirements into optimizable objective weights, guiding policy adaptation to diverse business preference. We conduct comprehensive evaluations of FLEXVNE on a simulation platform. Experimental results demonstrate that across topologies of varying scales and constraint intensities, FLEXVNE significantly outperforms all baseline schemes across all evaluation metrics. Furthermore, FLEXVNE successfully accommodates various user-defined intents, offering a promising paradigm for building intelligent and adaptive cloud-network resource orchestration systems.

The application of symbolic execution in the field of programmable network verification suffers from a lack of scalability due to path explosion and substantial computational resources consumption required for constraint solving. In this paper, we propose Scaver, a scalable verification system based on symbolic execution for programmable network. To address the path explosion problem, we develop multiple pruning strategies to reduce the number of execution paths without affecting the validity of verification, by precisely identifying the execution paths that are related to the verification purpose. To deal with the substantial computational resources consumption problem of constraint solving, we put forward an analytical results reuse strategy, avoiding repetitively computing the same constraints. Furthermore, an assertion language named PropLang is designed, which not only fulfills the demands of assertion language design, but also guides the implementation of symbolic execution optimization strategies. Experimental results on open-source programs demonstrate that Scaver is capable of effectively detecting potential network errors and alleviating scalability issue of symbolic execution. Compared to ASSERT-P4, Scaver reduces the number of execution paths and execution time up to 53%, 46%, respectively.

In the existing medical knowledge graphs, there are problems concerning inadequate knowledge discovery strategies and the use of single sources of medical data. Therefore, this paper proposed a research method for multi-data-source medical knowledge graphs based on the data, information, knowledge, and wisdom (DIKW) system to address these issues. Firstly, a reliable data source selection strategy was used to assign priorities to the data sources. Secondly, a two-step data fusion strategy was developed to effectively fuse the processed medical data, which is conducive to improving the quality of medical knowledge graphs. The proposed research method is for the design of a multi-data-source medical knowledge graph based on the DIKW system. The method was used to design a set of DIK three-layer knowledge graph architectures according to the DIKW system in line with the medical knowledge discovery strategy, employing a scientific method for expanding and updating knowledge at each level of the knowledge graph. Finally, question and answer experiments were used to compare the two different ways of constructing knowledge graphs, validating the effectiveness of the two-step data fusion strategy and the DIK three-layer knowledge graph.

The traffic scheduling mechanism in Time-Sensitive Networking (TSN) is the key to guaranteeing the deterministic transmission of traffic. However, when time-sensitive traffic and non-time-sensitive traffic are transmitted together, traffic scheduling conflicts are easy to occur in TSN. As a result, the deterministic transmission of time-sensitive traffic will be disrupted, and non-time-sensitive traffic may be preempted for a long time. To optimize the performance of multi-type hybrid traffic scheduling in TSN, we firstly establish a collaborative scheduling framework that incorporates Time Aware Shaping (TAS) and Cyclic Queuing and Forwarding (CQF) mechanisms. We then design a traffic shaping method in this framework based on Least Laxity First (LLF), which considers traffic characteristics to dynamically arrange the time slot injection sequence for different types of traffic. Finally, the traffic schedulability is evaluated based on the scheduling constraints of different types of traffic. Compared with the existing scheduling strategies, the proposed hybrid traffic scheduling strategy can schedule more non-time-sensitive traffic and achieve better delay performance of rate-constrained traffic in different hybrid traffic scenarios. When the number of flows is 100, the time slot injection ratio is increased by 24.3% compared with the LLF_TAS method.