How To Leverage Storage in the World of Hyperconverged Infrastructure
Recently, I wrote a blog post that attempted to answer the question about whether an organization must be 100% virtualized in order to use hyperconverged infrastructure. The answer to that question is ‘No’, but the article did open up a discussion on Twitter about the topic and turned into a discussion about how storage is leveraged in such environments. In this article, I plan to address the issue about how storage is consumed in the world of hyperconvergence.
A Bit of Background
Before I get started in a broader discussion, consider the various ways by which storage can be implemented and consumed in a data center:
- Direct-attached. Many places still use direct attached storage for some or all of their servers. This is storage that is in some way installed directly in a server and that stands alone. It’s not part of any grand storage scheme, but exists only to serve the will of the applications installed on the host server. For the purposes of the discussion here, don’t worry too much about typical single-server direct-attached storage.
- SAN/NAS. Still the most common deployment time today, SAN and NAS devices have for years provided organizations with plenty of capability and capacity to meet business needs. However, there are well-known frustrations in many of these kinds of environments, not the least of which is cost. For many, data storage is one of the largest infrastructure costs in the IT organization.
- Local, aggregated (hyperconverged). The storage in a hyperconverged environment consists of local (direct-attached) storage that is then aggregated with all of the storage in all of the other hosts in the hyperconverged cluster.[/unordered_list]
Storage Consumption In A Hyperconverged World
Moving to a hyperconverged infrastructure can often mean rethinking the way that you leverage storage in the data center. Most companies are not simply going to throw away their data center investments just to buy a bunch of hyperconverged appliances. As such, there will be a period of coexistence. Coexistence may even be a permanent state for those that choose to use hyperconverged tools for just part of their overall environment or that choose (or need) to maintain physical servers for certain applications.
With all of this in mind, there are two questions that often arise:
- Can you use the storage from the hyperconverged infrastructure in other parts of the environment? In other words, can a physical server consume storage from the hyperconverged infrastructure or can separate virtual hosts store virtual machine in the hyperconverged infrastructure?
- Can the hyperconverged infrastructure environment use traditional storage? That is, is it possible to store virtual machines on your legacy SAN while still running them inside the hyperconverged infrastructure?
First, the answers to these questions are technical only; they may not be considered best practice. After all, one of the reasons that companies buy hyperconverged is to simplify the storage experience. Running a virtual machine on one of the nodes in the hyperconverged infrastructure while storing it on a separate SAN negates much of this benefit. That said, there may be a need or desire to do something like this, so what’s actually possible.
Consuming Hyperconverged Storage From Outside the Environment
Suppose you have a regular vSphere host and want to use the storage inside your hyperconverged environment’s storage for that server. Can you do it? For these purposes of this article, I’ll lump VSAN in as a hyperconverged option since it’s the primary component that powers VMware’s EVO:RAIL.
It absolutely… depends. Different hyperconverged infrastructure vendors do things differently. VSAN, for example, does not allow external systems to access the hyperconverged storage. Neither does Nutanix. I’ve asked Nutanix why this is the case and the reasoning is this: the company wants to be able to guarantee and monitor storage performance and enabling the ability for external systems to siphon off IOPS runs contrary to this goal. Given that storage performance challenges are one of the reasons that companies choose hyperconverged infrastructure, this reasoning makes sense.
@DeepStorageNet @stu @otherscottlowe I was surprised by this answer seeing most solutions are just an NFS target.
— Keith Townsend (@virtualizedgeek) October 3, 2014
SimpliVity, on the other hand, has a different philosophy and does allow external systems to use the storage inside their environment. In my conversations there, the company indicates that they want to provide a more flexible overall experience with deep coexistence opportunity that can help more quickly solve storage challenges. So, that’s the way that they’ve designed their environment. Again, given the company’s goals, this reasoning also makes sense.
What you’ve seen here is the fact that different vendors provide different options for leveraging the storage that resides inside the environment. Whether the limitations are imposed because of technical hurdles or not, they are there for some companies. BUT, if you really wanted to use that storage trapped inside your VSAN or Nutanix environment, there’s no technical reason that you couldn’t create a virtual machine-based NFS server or iSCSI target and get at the storage that way. Of course, performance may be abysmal, but it would technically work.
Using Your SAN From The Hyperconverged Environment
With most hyperconverged storage options out there, there’s no reason that you can’t use your existing SAN or NAS device… unless your SAN or NAS device is of the Fibre Channel variety, in which case things might be a bit more difficult. In fact, there can be very good reasons to want to use your existing SAN with your hyperconverged environment. First and foremost, it makes transitioning virtual machine from the legacy environment to the hyperconverged environment far simpler. Just migrate the virtual machine and its disks from the SAN to the hyperconverged environment.
In most cases, you can still establish iSCSI, NFS, and SMB connections from whatever hypervisor is running in your hyperconverged environment (SMB for Hyper-V, of course). As such, the vSphere instances in the hyperconverged environment can, in fact, use your existing SAN.
Storage may or may not be handled a bit differently in the world of hyperconvergence. Check with your specific vendor to understand ways by which the hyperconverged storage can be leveraged from outside the environment and, where needed, make sure to understand that you can connect your hyperconverged environment to your non-FC SAN.