August 29, 2014 | Written by: Bin (Ben) Cao
Share this post:
Michael Fork guided you through building your OpenStack cloud at no charge using IBM Cloud Manager with OpenStack, an easy to deploy and use cloud management software with a 90-day download trial. Through the magic of the Chef-based deployer, you now have a brand new OpenStack cloud waiting to be used. Excited to get started, you might log onto Horizon, the web-based user interface for OpenStack services.
In addition to the community release (Icehouse) of OpenStack, IBM Cloud Manager with OpenStack is packed with unique features. For example, under the new Resource Scheduler tab, you will find the advanced IBM Platform Resource Scheduler (PRS). Let us take a look at how PRS will enhance your cloud experience.
PRS uses resource optimization placement policies instead of the default OpenStack compute scheduler filters. The placement policies allow you to make decisions on where to deploy the virtual machines (VMs) with specified resource selection policies. The optimization policy is adjusted dynamically. This means it is monitored both at VM deploy (VM creation) and during the VM life cycle. For example, when a new VM is deployed, the strategy monitoring system looks at where to place this VM based on the optimization policy selected. If a VM is deleted, then the monitoring system will also adjust accordingly based on the optimization policy; this could trigger a VM migration that rearranges the distribution of VMs across the VM hypervisors.
PRS supports various placement policies:
- The CPU Load Balance policy allows your OpenStack cloud to evenly allocate the CPU load across the host systems.
- The Memory Load Balance policy enables optimized distribution of memory load across the host systems.
- The Striping Policy evenly distributes the number of virtual machines across the host systems.
- The Packing Policy packs VM instances as densely as possible to maximize the number and density of running VM instances.
- The CPU Utilization Policy balances the CPU utilization among the host systems.
When selecting a policy, you will have a few policy attributes to manage. First, the Run Interval is the runtime interval in which the policy will be executed. In my example, I have selected a value of “1.” This means that the policy will be run every minute. The Stabilization is the number of consecutive times in which a policy goal can exceed the policy threshold value before the policy action is triggered. The Threshold is the starting point for live migration. Depending on the policy, this can be a percentage of either CPU utilization memory or memory utilization. Lastly, the Maximum Migration is the number of live migration operations allowed for the cluster.
In addition, you can customize the default behavior of Platform Resource Scheduler by creating your own plug-ins and then integrating them into Nova. To find out how to create your own plug-ins, or to learn more about PRS, go to IBM Knowledge Center to access the PRS documentation. Note that PRS is just one of the perks that come with IBM Cloud Manager with OpenStack, and you will have the option to use the default OpenStack compute scheduler filters in place of PRS.
I am interested in your thoughts about PRS or IBM Cloud Manager with OpenStack. Find me on Twitter @mrbencao.