Most hyper-converged infrastructure options create their own storage pools based on type of storage by grouping flash, performance disk drives and lower-performing media together. This boosts efficiency by ensuring workloads of higher importance get the storage performance they need, while non-production data can be stored on less expensive, lower-performing drives.
Interestingly, those storage pools are often presented back to the hypervisor as a NAS mount point. If they can leverage any shared storage resources at all, those resources are typically block or SAN storage resources.
There are some solutions that claim to be converged or hyper-converged that can leverage a NAS platform. Most of these solutions are what we at Storage Switzerland describe as pre-packaged solutions, but not converged. They are typically the three distinct pieces of the stack (compute, hypervisor, storage) that are often from three different vendors but are pre-integrated. These solutions reduce implementation time, but ongoing operations are similar to non-converged products. There are still three different operational interfaces, eliminating the simplified management benefit of hyper-converged products.
However, in 2014 we saw the introduction of two hyper-converged infrastructure products -- Atlantis USX and PernixData -- that started to better leverage existing shared infrastructure. These systems still aggregate the server’s internal storage but that storage then acts as an additional tier or cache in front of the shared storage resources. For some data centers this may be a more palatable way to move to a hyper-converged infrastructure. They can continue to leverage their investment in shared SAN or NAS while also taking advantage of the cost efficiency and performance advantages of server-side storage.
Hyper-converged infrastructures add support for more hypervisors
How hyper-converged options change storage provisioning