A Dynamic Programming Method for Data Center Server Resources

Abstract

The invention provides a method for dynamically planning server resources in a data center, which is suitable for configuring d servers in the data center. The data center includes multiple computer rooms, and each computer room includes multiple cabinets that can be configured with multiple servers, where d is a positive integer. This method first searches for cabinets whose remaining space is greater than or equal to d in all computer rooms to configure d servers. If the cabinet cannot be equipped with d servers, one of the computer rooms will be reassigned to a single computer room to free up a cabinet, so that the remaining space of this cabinet is greater than or equal to d, and d servers can be configured. If the cabinet cannot be vacated, reallocate across computer rooms to vacate the corresponding server locations of the corresponding cabinets in the computer room, so that the remaining space of these corresponding server locations is greater than or equal to d, so as to configure d servers.

Description

technical field [0001] The present invention relates to a method for planning server resources, and in particular to a method for dynamically planning server resources in a data center. Background technique [0002] In order to achieve economies of scale and provide tenants with nearly unlimited computing power and storage capacity, existing cloud data centers usually have tens of thousands of servers connected to each other through the network. And only a part of the servers and networks are provided to tenants for rent, so as to provide stable and high-performance cloud services to end users. Under the goal of data center operators to improve resource utilization in order to maximize the benefits of assets, high-efficiency resource planning technology has become a necessary key technology for cloud data centers. [0003] The cloud services built by each data center for tenants are various and vary in size, and the network bandwidth requirements of individual services are ...

AI Technical Summary

Problems solved by technology

However, the resource allocation mechanism of the existing data center to the network structure (topology) of individual cloud services is not a non-blocking (non-blocking) network, so in many cases even the network structure occupied by the service still has remaining bandwidth Wide, but may not be fully utilized because network traffic is blocked at some nodes

At the same time, the transmission paths of different cloud services may also interfere with each other due to shared network resources

Therefore, it is difficult for the existing mechanism to balance the three goals of data center resource utilization, network performance, and service stability

Moreover, the existing hybrid network resource sharing mechanism in the data center may also bring information security vulnerabilities

Regarding these challenges, there is still a lack of unified and mature solutions

Images