Browse Prior Art Database

System and Method for Workload Driven Dynamic Cloud

IP.com Disclosure Number: IPCOM000248835D
Publication Date: 2017-Jan-17
Document File: 6 page(s) / 78K

Publishing Venue

The IP.com Prior Art Database

Abstract

In a hybrid cloud (including physical machines, virtual machines and Linux containers) environment, most of the servers are served for one specified purpose: provide virtual machine service, bare metal service or container service. This is actually kind of “static cloud” which means that all of the servers are static and they can only provide one specified service. The problem is that in a cloud environment, the request for physical machine, virtual machine or container are dynamic, we cannot expect which kind of resources the user wants. If most of the request are container request, then all of the servers which can provide bare metal or virtual machine service may be idle but the servers which provide container service will be pretty busy. This will cause a low resource utilization in the cloud. This innovation was trying to make the cloud as a “dynamic cloud” based on workload. The “dynamic cloud” will switch cloud servers to provide different services based on workload. Such as the “dynamic cloud” can re-provision a server from virtual machine server to a container server etc. Making the cloud as a “dynamic cloud” based on workload can definitely increase the resource utilization for the whole cloud.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

1

System and Method for Workload Driven Dynamic Cloud

In a hybrid cloud (including physical machines, virtual machines and Linux containers) environment, most of the servers are served for one specified purpose: provide virtual machine service, bare metal service or container service. This is actually kind of “static cloud” which means that all of the servers are static and they can only provide one specified service.

The problem is that in a cloud environment, the request for physical machine, virtual machine or container are dynamic, we cannot expect which kind of resources the user wants. If most of the request are container request, then all of the servers which can provide bare metal or virtual machine service may be idle but the servers which provide container service will be pretty busy. This will cause a low resource utilization in the cloud.

The current solution for this is that when the customer wants to build a cloud, they need to have a clear request or blueprint for the cloud so as to decide how many different kind of servers that they need for their cloud. But this is not good enough as there may be future request change for the cloud, for such case, they will need to do some infrastructure update based on requirement which is not efficient. This innovation was trying to make the cloud as a “dynamic cloud” based on workload. The “dynamic cloud” will switch cloud servers to provide different services based on workload. Such as the “dynamic cloud” can re-provision a server from virtual machine server to a container server etc. Making the cloud as a “dynamic cloud” based on workload can definitely increase the resource utilization for the whole cloud.

It is actually kind of “Hypervisor Migration”. Most of the clouds are now trying to migrate some VMs or containers to balance the workload, but this innovation is trying to “Migrate Hypervisor” to balance the workload.

Before go through details for this invention, we can take a look at some key words.

Cold Server Group Hot Server Group

Threshold for Cold Server Group Threshold for Hot Server Group

Hot Server Group: The server group whose resource utilization is above the configured threshold and need to add some new servers to join this group to balance the current workload. Each host server group is composed by a specified kind of server.

Cold Server Group: The server group whose resource utilization is blow the configured threshold and need to switch some servers to join the hot server group to balance the workload in the hot server group. Same as hot server group, each cold server group is composed by a specified kind of server.

2

Threshold for hot server group: If the server group resource utilization is greater than this value, the server group will be treated as hot server group.

Threshold for cold server group: If the server group resource utilization is smaller than this value, the server group will be...