The problems to be studied or solved in this topic and the research methods (pathways) to be adopted:
I. Problems studied or solved
1. Complete the system architecture design and implement the following functions: based on the existing node attributes, the given parameters (such as maximum outdegree, maximum indegree, etc.) generate the edges of the connected nodes, so that the constructed graph is the workflow.
2. How to determine whether the condition for generating a directed acyclic graph is satisfied
3. Implement algorithms that enable structured DAG diagrams to generate multiple types of workflows for more accurate performance evaluation.
4. Provide an intuitive and effective reference model.
Second, the means to be adopted
1. Use the DAG generation algorithm togenerate a specific DAG map based on the existing node information.
2. Different generation strategies for different models
3. According to the figure to determine whether the requirements are met
Gradually demonstrate the abstract complex DAG graph topology sorting implementation steps, providing an intuitive and effective reference model.
Third, the subject overall (or software and hardware systems) solutions and research ideas
1. Using the theory and technical means, the construction method of the DAG diagram of the serial program task is obtained. First, select an appropriate granularity based on conditions such as processor execution rate, that is, the number of statements each node in the DAG graph contains. Secondly, for the loop structure in the program, the number of statements is calculated first. If it is larger than the selected granularity, it analyzes whether there are loop cross dependencies and loop independent dependencies, and converts its loop structure into DAG graphs. Again, convert the branch structure in the program to a node in the DAG graph. After analyzing the loop and branch structure, the program is divided into blocks according to the selected granularity. Finally, the sequence structure is analyzed by the algorithm to generate a complete program DAG graph.
2. Directed acyclic graphs (DAGs) can reflect the mutual constraints between tasks in the workflow and are therefore widely used to discuss workflow scheduling problems for parallel task clusters. However, with the explosive growth of the number of mobile applications and cluster tasks, the existing single DAG workflow-based scheduling scheme has the following problems: (1) It is impossible to balance all situations and comprehensive situations in multiple scenarios of multiple DAG workflow complex scenarios. Considering the overall energy-saving effect; (2) the performance optimization effect of multi-DAG workflow is acceptable, but it brings a lot of energy consumption, and can not balance performance improvement and energy consumption optimization; (3) the application scenario has greater limitations, the algorithm Poor fitness. I will propose an energy-saving scheduling algorithm based on a multi-DAG workflow that can solve the above problems and reduce energy consumption and improve work performance.
3. Communicate and communicate with industry peers and companies, the public and other users on complex engineering problems encountered in workflow generator design, write phased R&D reports and design contributions, clearly express demand analysis conclusions and respond to user instructions correctly.
4. Regular communication, continuous improvement of models and optimization algorithms to improve the usability of the system (or method)