Research On Task Management And Resource Allocation Mechanism For Large-scale Heterogeneous Evaluation Scenarios

As Internet technology continues to mature,online education has also grown vigorously.Especially in the field of IT programming,due to the wide variety of IT technologies and its characteristics of rapid replacement,the online education mode of on-demand teaching just caters to the deep needs of IT professionals.The computer field is a typical engineering field.Online education in text or video form lacks practicality,what students learn in the class cannot be effectively absorbed and it is difficult to achieve the learning effect.The existing online coding platforms can only provide some online programming and assessment of basic algorithm,but cannot provide support for various advanced technologies such as web frameworks,machine learning,distributed cloud computing architectures,which are commonly used in enterprises.Programming education is currently disconnected from corporate needs.With the development of virtualization technology,some virtual platforms have begun to provide virtual practice environments for various IT technologies,to simulate real enterprise application scenarios.However,a problem still existing on these platforms is that,there is no general assessment mechanisms,thus students cannot be quantitatively measured of the learning effects and given corresponding guidance.In view of the current shortage of online programming education,we proposed an evaluation cloud platform that provides teaching and evaluation for large-scale heterogeneous IT technologies.Based on the evaluation of the existing task management and resource allocation strategies of the cloud platform,this paper proposes an improved task management and resource allocation mechanism and integrates it with the evaluation cloud platform to verify the feasibility and effectiveness.The main contribution of the article is:First,a task management mechanism of large-scale evaluation cloud platform based on evaluation task’s characteristics.We first proposed an improved general evaluation task management mechanism for existing single-point bottlenecks and inefficiencies in current task management mechanism;at the same time,in order to ensure the real-time evaluation and effective evaluation cluster resources utilizing,for simple types of tasks,we propose a task container reuse task management mechanism;in addition,for the support of engineering-level training tasks,this paper studies and proposes a container web terminal and web virtual desktop mechanism based on the Web Socket protocol.Second,a container resource allocation mechanism based on intelligent analysis of runtime data.The original manual resource allocation method based on human experience may cause low resource utilization or insufficient resource allocation.In order to improve the utilization of task container’s computing resources,based on the evaluation of the massive historical operating data of the cloud platform,this paper analyzed it with the help of machine learning methods,and established the relationship between the complexity of an training itself and the computing resources required for the training.And we verified the superiority of our method through experiments and analysis.Third,an improved large-scale evaluation cloud platform.Based on our proposed task management and resource allocation mechanism,we designed and improved the key modules of the evaluation cloud platform.The task scenarios for evaluating cloud platforms can be extended to typical cloud platform task scenarios,and the related task management mechanism research and resource allocation work have certain reference significance in similar cloud platform construction.