Browse Prior Art Database

Dynamic Scaling of host groups in Cloud infrastructure

IP.com Disclosure Number: IPCOM000249138D
Publication Date: 2017-Feb-08
Document File: 4 page(s) / 331K

Publishing Venue

The IP.com Prior Art Database

Abstract

The following article tries to address the mentioned issues by dynamically restructuring the host group to optimize it with policies that specifically target to bring energy efficiencies and dynamic scaling to cope up with cloud demand.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 34% of the total text.

1

Dynamic Scaling of host groups in Cloud infrastructure

Problem Statement In Cloud deployments, the capacity usage tends to get skewed w.r.t. hosts that are hosting Virtual Machines(VMs). Over time, there are two situations that could arise:

1. Some hosts have VMs which are not using its resources to full capacity, resulting in low energy utilization.

2. Some hosts have VMs which are running with ful utilization of host’s resources. In recent cloud deployments, many use the strategy to group hosts. For private clouds, a set of hosts is then used for a specific organizations within a company. For ex. a host‐group could be assigned for hosting VMs for development within IT unit of the organization. Similarly, another set of hosts could be assigned for test purposes. The probability of under and over utilization of hosts gets increased when you take into consideration hosts subsets.

Solution Overview and Details There are many cloud deployments that have the notion of grouping hosts. An example is OpenStack. OpenStack provides to the cloud user a way to group hosts. A host group can then be assigned to a development/test IT unit of an organization. For organizations following the waterfall model, there will be times when one host group will have Virtual Machines that are heavily utilizing host group resources when compared to another host group. This is because in Waterfall model, the development unit has higher use of VMs in the first stage, and the test unit has the higher use of VMs in later stage. Waterfall model is one example; this situation can arise in different scenarios such as user consumption model, geographical IT units etc. In such situations, the cloud infrastructure is unbalanced. To address this, one has to target the border cases:

1. Balance the host group for energy efficiency 2. Scale out host group based on its demands.

The main feature of this article is to define an interconnect policy that can balance host groups by leveraging the under‐utilization and over‐utilization of host groups. The article is define a host group that will act as a pool of unused hosts or pool with hosts ready to be used, let’s call it as “ReadyToUse” host group. In the following sections, we will define how this “ReadyToUse” pool of hosts gets used in re‐balancing the cloud infrastructure which has host groups defined.

2

Balance the host group for energy effeciency

The following can arise in host group 1. There are less number of VMs deployed on one/some host(s). 2. The VMs have very minimal utilization on one/some hosts(s). 3. Etc

In such cases, an energy efficient policy can be defined on the host group. This policy will primarily dictate on how to move the VMs to different hosts of the group so as to get a set of hosts that can be switched off. The policy needs to take into consideration the following:

1. Define a value for average utilization of host. 2. Define a selection criteria to choose a VM that can be moved to ...