A computational node cluster elastic distribution system and method

Abstract

The invention relates to a computational node cluster elastic distribution system and method. A computational node cluster elastic distribution mechanism is adopted, according to the usage of computing node resources and task resource requirements of historical tasks, the resource allocation of computing node is predicted, and under the condition of satisfying the demand, the allocation of computing phase is dynamically controlled, the job response speed and the resource utilization rate of computing node are improved, and the running results of historical prediction will be fed back to the next prediction. Therefore, the balanced allocation of resources in the cloud computing environment is realized and the comprehensive efficiency of the system is improved.

Description

technical field [0001] The invention relates to the field of dynamic allocation and management of cloud computing cluster server resources, in particular to a computing node cluster elastic allocation system and method. Background technique [0002] Along with the development of the computer field, the development of the cloud computing field is particularly rapid. Cloud computing provides users with convenient, fast and safe data storage and network services through computer and network technologies such as distributed computing technology, parallel computing technology, virtualization technology and load balancing. [0003] Most of the operations involved in deep learning are vectorized matrix operations. As a graphics accelerator, the GPU provides a large number of computing cores for rendering. These computing cores can also be used to accelerate vectorized matrix operations. Therefore, in recent years, deep learning uses GPUs for model training. As demand increases, m...

AI Technical Summary

Problems solved by technology

[0004] However, due to the particularity of computing node resources in hardware, cloud computing node resources are usually provided to users in an exclusive manner, and this allocation is one-way and static, which can easily cause computing node resource overload and poor user experience. good things happen

[0005] In the exclusive mode, each computing node resource is difficult to maximize the performance

The fixed resource allocation method makes it difficult to efficiently match the needs of different tasks requested by different users

In the case of only one allocation of computing node resources, when the assigned user actually submits the task and starts computing, the initially allocated computing node resources may not be able to meet the user's computing needs

Images