QBT: Efficient and Flexible Resource Allocation Method for Data Center of State Grid Scenario

Abstract

With the rapid development of computer technology, service platforms have increasingly become the industry's first choice for building service center. In order to use resources more efficiently, flexible resource allocation is a very important part of the service platform construction. In the service of State Grid, the issue of flexible resource allocation is super important. However, the business of the State Grid is very complicated, and the service will include the invocation of various business logics, and resource allocation cannot be simply carried out. In order to solve this problem, this paper designs an efficient and flexible resource allocation scheme-QBT (quota boosting task) based on this project. QBT has designed three functional modules: log collection, log analysis, and flexible resource allocation. Log collection and log analysis will restore the real request path and construct it as a DAG. The system will count the QPS of each DAG, and calculate the load corresponding to the service on the DAG. Based on this information, the elastic resource allocation will perform elastic resource allocation in the two situations when the remaining resources are sufficient and insufficient. The experimental results show that the QBT design scheme can successfully solve the resource scheduling problem of the National Grid Service Platform.