For some time I have been saying that a real IT shop would be ill advised to attempt consolidation to a single hypervisior. Why?
My sense is that we have two types of workload, the mission critical, drives the business applications and services that cannot fail. These are the Fedex Logistics platform, the eBay portal, the British Airways automation system, Citibanks credit card processing systems. These applications make up about 15% of the total application manifest and are different from the rest. We have always mollycoddled them, clusters solutions, disaster recovery plans and platforms, extra vigilance.
For the mass of real business applications the idea of multiple hypervisors maybe isn’t necessary but for those – you get fired for dropping them – applications. I bet most CIOs will take the risk adverse route. That leaves operations with the complexity to manage.
If you have Oracle applications, then Oracle VM is the risk adverse, least pain choice for the CIO. If you want virtualization in the Windows world, then you ought to use Hyper V as that is the only platform that Microsoft will support unreservedly. You probably have a large VMware environment already containing the Windows sprawl we all started to create last century. Maybe you need Citrix and KVM as well?
This might sound like a complete nightmare, with many different infrastructure platforms, different operating procedures and varying levels of maturity and capability. Actually it sounds a bit like last century when we had Solaris, Aix, HPUX, Windows and Linux to manage. We just slipped down a notch to the hypervisor as the point of differentiation.
There is some good news – we can use the same hardware platforms (mostly)
We can use the same network, but may have challenges around managing bandwidth between hypervisors as current virtualization technology serves IO weakly.
Can we use the same storage? As for network, shared storage across multiple hypervisors could be a problem in managing workload, avoiding hotspots and providing a degree of isolation between systems to avoid the risk of conflicts and interactions.
Storage controllers that offer automated tiering capability and multi-tenancy are likely to provide that level of isolation between hypervisor stacks that makes up for non existent IO handling and management between hypervisors.
Yesterday I met with Jay Kidd at NetApp and we talked at length about the cloud and the place of storage in the cloud. NetApp have been working with Systems Integrators for some time and have built up a great deal of expertise in this area.
Here is what my colleagues Terri McLure and Mark Peters have to say about NetApp Secure Multi-Tenancy:
“Using MultiStore functionality, users can logically partition storage system, network and disk capacity. MultiStore has been shipping for a number of years; it is what NetApp leverages in its vFiler storage virtualization solution to enable shared storage pools.”
Using multi-tenancy capability sounds like an ISP or Systems Integrator solution but I would argue that this capability is now ideal for the enterprise who needs to support multiple hypervisors.