What do we want?

Management-Speak

"A highly available virtualisation pool that copes with hardware failure and service interruption, integrating aspects of autonomic decision-making, using configuration management for both service and guest deployment, whilst remaining performant and cost-effective in comparison to physical hardware solutions."

Availability

• Guests should always be running
• Migrations should be live
• No scheduled downtime

Redundancy

• Hosts go down
• Disks fail
• Guests crash

Flexibility

• Load balancing, monitoring, scheduling
• Scaling must be easy
• Automate as much as possible
• Manual control available if needed

How do we get it?

Virtualisation

Libvirt & qemu/KVM

• Handles live migrations
• Hugely configurable
• Guest OS agnostic

Storage

(rbd)

• Network block devices
• Distributed & replicated
• Scaleable (automatically)
• Sparse disk images
• qemu integration
• ...

Ceph in Action

Cluster Management

• 'Cloud Management Platform'
• Controls Resources:
• VM Hosts
• Disk Images
• Guest VMs
• Guest networking
• Done with libvirt & ceph integration

Useful Features

• Relatively simple to set up
• Resource creation from templates
• Failure monitoring & hooks
• Load-balancing & guest placement
• Can be turned off

OpenNebula in action

Management Tools

OpenNebula Sunstone

Command-line

• Similar functionality to Sunstone
• Scriptable, parseable
• 'Personal preference'

LCFG

• Component:
• Resource creation (& deletion)
• ONEDB backup and restore methods
• External inventory integration

Demo (?)

opennebula

LCFG

vmhost 1

vmhost 2

guest VM