Section 14: Multi Node Design and Scaling Openstack

51. Scaling Openstack

https://www.udemy.com/course/openstack/learn/lecture/6800910#overview

• Walmart runs on Openstack
• Where to put what?

Scaling Openstack

• Horizontal Scaling
  • Scale by adding more machines into your pool
• Vertical Scaling
  • Scale by adding more power to existing machines
    • CPU, RAM

How Openstack Components Scale

• All components use REST
• Load Balance using HA Proxy using a single endpoint
• Databases do not scale naturally
  • Can use Active/Passive
  • Instead use Galera for clustering.
• Message Queue has it’s own clustering method
  • RabbitMQ, not RappidMQ
• Non-API components offer built in support for scaling
  • Neutron, swift, proxy, L3

52. Compute Node

https://www.udemy.com/course/openstack/learn/lecture/6809170#overview

Compute node requires

• Services
  • Nova compute
  • Neutron
  • Hypervisor
• Compute
  • High CPU and RAM
• Storage
  • External
  • Distributed
  • Direct
    • Fast/performance
    • Loss of data = no return

53. Network Node

https://www.udemy.com/course/openstack/learn/lecture/6809800#overview

54. Storage Node

https://www.udemy.com/course/openstack/learn/lecture/6809822#overview

• Different types of storage should be on different nodes.

55. Controller Node

https://www.udemy.com/course/openstack/learn/lecture/6810094#overview

• provides front door.
• Most API endpoints run on Controller

56. Multi Node Design & Minimum Node Requirements

https://www.udemy.com/course/openstack/learn/lecture/6810186#overview

• 2 node
  • controller
  • compute
• 3 node
  • controller
  • compute
  • network
• 4 node
  • controller
  • compute
  • network
  • storage
    • Glance

Node Requirements

add image

Section 15: Expanding the Cluster: Adding compute

57. Preparing the node