Web-scale architectures are rapidly making their way into the core of enterprise datacenters. This shift is being driven by Nutanix. One of the keys to web-scale is the idea of a shared-nothing architecture. This means that there can be no single point of failure, no bottlenecks anywhere in the system, and scalability with no fixed limits.
One thing Nutanix engineers realized early on is that adherence to these rules demands that our Nutanix Operating System (NOS) not rely on external components – including management consoles such as vCenter. This was a significant decision given vCenter’s central role in the majority of virtualized datacenters. Make no mistake, we expect our customers to use vCenter to manage their VMware clusters on Nutanix. We just made sure that vCenter works seamlessly with NOS, while not being dependent upon it.
Each Nutanix node communicates directly with the host hypervisor. This communication doesn’t depend on the network, and doesn’t require an external management entity like vCenter. vCenter can be deployed to the Nutanix datastore, and then configured to manage the cluster on which it runs. This means that no alternative infrastructure is required to run the vCenter VMs. Also, if vCenter is unavailable for some reason the Nutanix filesystem will not be impacted. It is treated like any other VM running on the hypervisor – not dependent on vCenter to operate.
Imagine if Nutanix required vCenter to configure or start the Nutanix cluster. For one thing, this would require every Nutanix customer to have some alternative infrastructure to run their vCenter instance, and it would make ROBO or tactical deployments considerably more complex and less robust. Now contemplate if Nutanix somehow depended on vCenter in order to maintain operations. In this case, vCenter and the hardware on which it runs becomes the most critical part of your infrastructure. Your VM availability is only be as good as vCenter’s availability. Consequently, all of the reliability benefits of implementing a distributed architecture begin to vanish, as your entire infrastructure now depends on the same legacy infrastructure that forward-thinking customers are trying to escape.
Additionally, consider how difficult it would be to support alternative hypervisors, such as Microsoft Hyper-V or KVM, if the management tool was a dependency. KVM for example doesn’t even have a standard management tool, and Microsoft System Center Virtual Machine Manager (SCVMM) has entirely different APIs than vCenter. The fact that only hypervisor communications needed to be abstracted enabled our engineers to quickly support these other hypervisors, and to re-use the vast majority of our existing code.
Nutanix is equally robust and feature-rich across all hypervisor platforms. In other words, it’s hypervisor agnostic and does not burden customers with having to deploy adjacent infrastructure to host vCenter. The result is greater resilience and more choice.