Multi-HV OpenStack - It's 2015 already, are we there yet?

Editor's Notes

• #5: Who was on our previous talk in Paris? Who’s using VMWare? Who’s using Oracle VM or Xen? Who’s integrating the two into OpenStack? Who would like to be able to?
• #6: Who was on our previous talk in Paris? Who’s using VMWare? Who’s using Oracle VM or Xen? Who’s integrating the two into OpenStack? Who would like to be able to?
• #7: So here’s the first one - developer enablement. Mainly it means that some organization which happily runs VMW-based infrastructure (vSphere+vCenter, maybe vCOPs next to them) wants to “jump on” the DevOps/Agile/whatever (else buzzword) train and enable the developers in the organization with cloud-based capabilities - rapid self-service infrastructure provisioning, application deployment orchestration, DevOps-friendly APIs. All this is usually done under “reduce time to production” flag, and OpenStack is perfectly capable to deliver this - in combination with lower-level resources (VMs, networks, block devices) powered by VMWare The second one is what we agreed to call Infrastructure optimization. The mission of OpenStack is to enable massive user choice, and with OpenStack an organization can easily combine a number of hypervisor, storage and networking technologies under single end-user API. In such case, VMW-based infrastructure becomes one of the underlying resources exposed through OpenStack, along with open-source bits like KVM or Ceph, Oracle-certified configs on Oracle VM and etc. Heterogeneous infra – avoid vendor lock in On the last one I’d like to stop a bit more
• #8: 0.5 minute So here’s the first one - developer enablement. Mainly it means that some organization which happily runs VMW-based infrastructure (vSphere+vCenter, maybe vCOPs next to them) wants to “jump on” the DevOps/Agile/whatever (else buzzword) train and enable the developers in the organization with cloud-based capabilities - rapid self-service infrastructure provisioning, application deployment orchestration, DevOps-friendly APIs. All this is usually done under “reduce time to production” flag, and OpenStack is perfectly capable to deliver this - in combination with lower-level resources (VMs, networks, block devices) powered by VMWare
• #10: We’ll focus on KVM + vSphere Containers – you had enough of that already 
• #12: So here’s the picture quite similar to what you might have seen already. VMWare folks have done GREAT job developing drivers to make sure they technologies work the best possible way with OpenStack. On Compute side, we get a vCenter driver that allows Nova to place VM provisioning requests on ESXi clusters via vCenter On Block Storage side, we get a VMDK driver which allows us to create Cinder Volumes on whatever storage technology (VSAN, EMC, etc) is exposed behind VMDK - and attach these to virtual machines. Quite performant, btw, no iSCSI overhead On Image/Snapshot storage side, we get a driver which also hooks to VMDK to store VM images and snapshots there Finally, on Networking side, we connect OpenStack’s Neutron to VMWare NSX as an SDN backend to power networking of this cloud with all great stuff which programmable networks deliver
• #13: 1 minute Now, under another perspective, here’s how the cloud with combination of ESXi clusters-as-hypervisors and KVM nodes look like. So that’s the idea that we tend to call “heterogeneous cloud” - an OpenStack cloud containing various types of underlying resources (in our case - hypervisors best suited for hosting different kinds of applications) “united” under single OpenStack umbrella. I becomes up to cloud owner how to expose these to users - as different Host Aggregates, Availability Zones or specific VM image types.
• #14: 1 minute VMware’s HA/FT - protect your virtual machines from failures, by enabling HA on ESXi cluster (rapid restart of a VM in case of failure) or FT (zero-downtime VM failover in case of failure) DRS - dynamic balancing of VM placement within ESXi cluster vMotion - zero-downtime live migration of VM NSX - rapid self-service creation of virtualized networking topologis (incl. advanced services such as firewalls, load balancers) for applications, insightful networking troubleshooting tools This all works for VMs provisioned from OpenStack into VMware infrastructure :)
• #17: So here’s the picture quite similar to what you might have seen already. VMWare folks have done GREAT job developing drivers to make sure they technologies work the best possible way with OpenStack. On Compute side, we get a vCenter driver that allows Nova to place VM provisioning requests on ESXi clusters via vCenter On Block Storage side, we get a VMDK driver which allows us to create Cinder Volumes on whatever storage technology (VSAN, EMC, etc) is exposed behind VMDK - and attach these to virtual machines. Quite performant, btw, no iSCSI overhead On Image/Snapshot storage side, we get a driver which also hooks to VMDK to store VM images and snapshots there Finally, on Networking side, we connect OpenStack’s Neutron to VMWare NSX as an SDN backend to power networking of this cloud with all great stuff which programmable networks deliver
• #23: 0.5 minute Now, previously I had a slide with Fry from Futurama screaming “shut up and take my money here”, but it was not flying well with people, kinda turns the whole story into a sales pitch :) So just to answer the question - how to build a cloud with such type of integration?
• #24: 1 minute So today there’s a number of options out there how one can get a VMWare-integrated OpenStack cloud running. #1 is obviously VMWs own offering - VIO, or VMWare Integrated OpenStack. Mirantis OpenStack with Fuel for deployment also supports a number of use cases with VMWare bits integrated. And finally, for those of you feeling adventurous, there’s an option to do a DIY config - after all, the drivers are up there in Community trunk :)