This post is very different that I have even posted. Not sure If I will continue such musing. Both this is one of those thoughts kept lingering in my mind for a while.
By Multi-Hypervisor environment I meant using more than Hypervisor.
1. vSphere ESXi
3. RHEV/XEN (Not sure if this is the name)
Disclaimer: I know only one Hypervisor i.e. VMware’s vSphere ESXi
First why Multiple Hypervisor’s
1. Single Vendor/All Eggs in one Basket
Historically CIO’s never want to dependent on Single Vendor. This is goes back to same strategy used in olden days for hardware vendor. Big Enterprises will always have mix of hardware Vendor and Strategy e.g. High End systems from HP and Mid and Low End system from IBM & Dell(or other way round). In short use multiple server vendors and remove any possibilities of any vendor to implicitly rule organization’s hardware strategy.
2. License, Support and Ecosystems
Licensing cost will be one of the factor but not the only factor while deciding one Hypervisor or the other. I feel the major influence will be the service which are sold to the internal client or how the charge back model is gets impacted by the overall cost. I think Hyper-V can definitely provide Silver, Bronze level service or uptime and vSphere can do that right from top (Platinum) to bottom (Bronze). But then will Silver level service cost same on vSphere and Hyper-V. I don’t know the figures but we know cost differs by significant margins. CIO’s must direct Infrastructure architects to develop multiple hypervisor strategy and service models
Support and Ecosystem becomes extremely critical in ever becoming complex environment. Ecosystem I meant your configuration and monitoring software. Both Vendors are building up capability to support multi hypervisor environment. For me it is always looks like Indian marrying a Indian Born in UK. Bottom line –> There will be always a case where Vendor A will point to Hypervisor-B and Vendor B will point to Hypervisor A when the problem arises in troubleshooting cross vendor hypervisor software. I would always recommend to use software’s of the respective vendor to management, monitor and configure the ecosystem.
Impact of Multiple Hypervisor’s
1. Evolving 3rd Party Software
3rd Party Software vendor’s started developing product to support multi-Hypervisor environment. I know at least Veeam is doing it. And there might be many but I see major change happening very soon. Right now let’s say we have Gold, Platinum VMs in vSphere ESXi and Silver, Bronze are in Hyper-V, there will be time when customer wants to upgrade to Gold or Platinum service or in simple words customer felt vSphere will offer better stability so they want to move to vSphere i.e. at lower level it is VHD to VMDK conversion. No…not offline but online conversion. If there is Storage vmotion why can’t be vmotion irrespective of hypervisor. After all now we aim everything to be defined by the Software. I have always believed it is not the need that pushes to innovate but the innovation drives the need. e.g. Apple iPhone. World was happy in pre-iPhone world but now everyone wants iPhone and every other phone started copying iPhone interface.
So I see this segment will emerge. Platspin was number 1 vendor when it came to P2V and capacity planning for P2V, I see similar kind of software will evolve where in online conversion happens from VHD to VMDK or vMotion/Live Migration happens over different Hypervisor. CIO’s will always love such flexibilities.
2. Complexities and System Administrators
Biggest impact Multiple hypervisor will have is on Operations team. Supporting multi-hypervisor needs training and of course understanding the subtle differences. Different maintenance cycle and upgrade procedures and that too without impacting services will need detailed strategy and planning. Even with single hypervisor I see operations team (network, Storage and OS) acting independently in their own Island of excellence (Is it). If these team don’t talk frequently there is always a chance for major outage. I think operation team must be renamed a datacenter operations team and whole team becomes responsible for virtual datacenter. If Network is down Hypervisor is down, so are the services and if you have FCoE even Storage then. If they don’t work closely such outage window can become very long. Entire team ,ust understand the impact of services when two ports on different switch goes down. This has potential to bring 50-100 services down and business loss.