What is the future of virtual storage in a Software Defined Data Center (SDDC)? As more and more technology gets moved from hardware to software in the SDDC, I have to wonder which direction virtual storage will go.
If we use networking as an example, we see that the technology has evolved from local virtual switches on each of the hosts to a virtual distributed switch (VDS) model where all the individual host-level virtual switches are abstracted into a single large VDS that spans multiple hosts at the data center level. In this design, the data plane remains local to each VDS, but the management plane is centralized, with VMware vCenter Server acting as the control point for all configured VDS instances.
It would seem logical that virtual storage will follow a similar path and would be presented in some sort of distributed model. VMware has moved forward with that idea in their version of Distributed Storage (DS). DS is a distributed layer of software running natively as part of the ESX hypervisor. It aggregates the hosts’ local storage devices (SSD and HDD) and makes them appear as a single pool of storage shared across all hosts. That should sound familiar, in that what they are doing with local storage is what they have done in the past with CPU and memory – virtualize the physical resources of ESX hosts and turn them into pools that can be carved up and assigned to VMs and applications according to their QoS requirements.
That sounds really great and seems to be moving in the right direction, but when you look at how virtualization has evolved over the years, the concept of using individual host resources takes me back to early days of VMware 2.X when Virtual Center was released and the hosts became clusters. In my humble opinion, this shows that DS is either very much in its infancy or is switching directions to match and build on what Microsoft’s Hyper-V has been doing for a little while now. VMware’s Virtual Storage Appliance was the first step on the way to DS that worked with small environments, where of the biggest drawbacks was scalability into larger environments. It appears that drawback has been resolved
When I hear the term distributed storage, I actually picture something a little different. I am thinking mainly of fibre SANs for the backend storage and fibre switches virtualized directly into the hypervisor. This would present the ability to perform the WWN zoning from inside the hypervisor. EMC has already demoed a virtualized Isilon SAN running on vSphere; this shows that the service processer of the SAN can be virtualized in the same way virtual software SAN companies like Datacore and Falconstor have been doing already. That gives us the ability to zone and mask the storage from inside the hypervisor, but I do have an initial question: How would you recover from a complete data center crash?
In conclusion, VMware’s direction forward with their DS is not quite the direction I thought and imagined it would be, especially where EMC, the parent company, also likes to sell disk subsystems. The evolution of the VSA and DS does make complete sense to me, but Microsoft has been utilizing this method with Hyper-V for a while now, and it does appear they have had the right idea.