Grounded in the Cloud
Showing results for 
Search instead for 
Do you mean 

How to improve VM creation time with HPE Helion CloudSystem in an OpenStack Environment

HaleyC on ‎10-05-2016 01:56 PM

Post by Guest Authors, Siva Subramaniam HPE Cloud Scale and Performance team and Thavamaniraja Sakthivel - HPE Cloud Scale and Performance team

In Helion CloudSystem, when large number of computes are registered to cloud (>10 computes), the Virtual machine boot time may degrade when there are large number of concurrent requests. This is because virtual machine provisioning gets scheduled in the hypervisors in stacking mode by default. In stacking mode, VMs are provisioned in set of 5 compute nodes. The scheduler picks up next set of 5 computes only when the first set of computes is completely exhausted, this in turn increases the boot time for the virtual machine when the concurrency of nova boot request is more than 5.

To improve the VM boot time in a large scale environment, scheduler configuration can be modified from stacking mode to spreading mode. In the case of Spreading Mode, the VMs are provisioned on the computes where the largest resources are available. For example, in a cloud with 20 identical compute nodes, and if there are 20 concurrent boot requests, the scheduler schedules the VM creation on all 20 compute nodes in parallel as against stacking mode. This improves the overall boot time of VMs.

 To modify the scheduler configuration to spreading instead of stacking, follow these steps:

  • SSH to deployer appliance
  • Open the file ~/helion/my_cloud/config/nova/nova.conf.j2
  • Modify the parameter ‘ram_weight_multiplier’ to 1.0

Run the following commands for the changes to take effect in the cloud:

  • cd ~/helion/hos/ansible
  • git add –A
  • git commit –m “commit message”
  • ansible-playbook –i hosts/localhost config-processor-run.yml
  • ansible-playbook –i hosts/localhost ready-deployment.yml
  • cd ~/scratch/ansible/next/hos/ansible/

ansible-playbook –i hosts/verb_hosts nova-reconfigure.yml

VM boot comparison results between stacking and spreading mode:

Configuration:

Number of Computes: 20 with 256 GB RAM

Number of concurrent request: 20

Stacking Mode:

CStech1.png

As shown in the result above, Average boot time of 20 instances with 20 computes nodes in stacking mode takes 61.612 sec.

Spreading Mode:

tech2.png

Whereas with the spreading mode (as shown above), it takes 21.702 sec on average for the same number of concurrent request with 20 compute nodes. 

For more information on case studies, demo trials, support, technical information, and more check out the HPE Helion CloudSystem library

 

0 Kudos
About the Author

HaleyC

I manage the HPE Helion social media brand accounts promoting the enterprise cloud solutions at HPE for hybrid, public, and private clouds.I have put my toes in the ocean of cloud evangelism for the enterprise IT industry. But my expertise is in Social Media and Digital Marketing.

Events
June 6 - 8, 2017
Las Vegas, Nevada
Discover 2017 Las Vegas
Join us for HPE Discover 2017 in Las Vegas. The event will be held at the Venetian | Palazzo from June 6-8, 2017.
Read more
Apr 18, 2017
Houston, TX
HPE Tech Days - 2017
Follow a group of tech bloggers for a new HPE Tech Day, a full day of sessions about how to create a hybrid IT, from hyperconverged to Composable Infr...
Read more
View all
//Add this to "OnDomLoad" event