VMware has joined the ranks of the hyperconverged, announcing VMware EVO:RAIL, a solution built on its vSphere foundation coupled with VMware VSAN. Readers here know that the converged infrastructure available for years now solves some problems in IT, but frankly, most of those problems are political. The guts of converged infrastructures are largely business as usual, have many of the same needs and time sinks as traditional infrastructure, and comprise largely the same inflexible hardware you’ve always used.
Hyperconverged infrastructures, such as those from Nutanix, SimpliVity, Scale, and now VMware, look to change the way IT does business by fundamentally changing the technology in use. They take commodity compute and storage hardware and apply “software-defined” techniques to it to drastically reduce the time and complexity of provisioning and ongoing maintenance. Most are delivered as preconfigured hardware, and VMware EVO:RAIL is no exception. The basic building blocks are similar to those of other hyperconverged offerings: 2U containers VMware calls “Hyper-Converged Infrastructure Appliances” (HCIAs), which contain up to four individual compute and storage nodes. Partners participating in this program will need to follow a strict “recipe” for the hardware, using specific types and sizes and speeds of CPUs, RAM, storage, and networking. Practically speaking, I’d expect that nearly every EVO:RAIL will be a Dell PowerEdge C6220 II or a Supermicro Twin, as those two companies are the only ones left in the modern x86 hardware game anymore.
VMware EVO:RAIL is built upon VMware vSphere, the company’s solid foundational hypervisor and management tool suite. It adds VMware Virtual SAN technology to provide storage, as well as the custom EVO:RAIL engine for automation. Use of these technologies informs how large a single VMware EVO:RAIL cluster can be, topping out at only four HCIAs, for a total of sixteen nodes. Each HCIA, or set of four nodes, is intended to serve 100 general-purpose workloads (2 vCPU, 6 GB memory, 60 GB disk) or 250 virtual desktops (2 vCPU, 2 GB memory, 30 GB disk). A group of 400 general-purpose VMs, or 1,000 virtual desktops, is well-suited for VMware’s target audience in the remote-office, VDI, and virtual private cloud markets, and it fits nicely with the industry trend of looking to serve the millions of smaller organizations out there that need IT but don’t have the resources to manage their IT well.
The VMware EVO:RAIL engine aims to allow organizations to configure their systems in just minutes, manage them without the traditional interface complexity, and perform nondisruptive operations like upgrades, patches, and scale-out activities. The interface is pure HTML5, which is a refreshing choice because it means widespread browser compatibility and fewer support headaches.
So, Is VMware EVO:RAIL Worth It?
When I look at this new solution, it seems to me that VMware is attempting to arm its partners against the entrenched hyperconverged vendors like Nutanix and SimpliVity. That’ll likely muddy the waters some, but version 1.0 of EVO:RAIL is missing a number of features, like replication and deduplication, that are essential to many customers. It would also be nice if EVO:RAIL would scale up further. Nutanix can go to forty-eight or more nodes, for example, not just sixteen, and you can customize the types of nodes to account for your workloads and storage needs. This won’t be easy to achieve without fundamental changes in the whole vSphere stack, though.
EVO:RAIL doesn’t solve problems like backups, either, so customers will need to delve into the underlying complexity if they wish to protect their data (and who doesn’t?) Like all other hyperconverged solutions, it also completely ignores networking. You have to supply your own 10 Gbps network switches, and with that you’ll need a network engineer to configure and operate it. It is a sad truth that most IT folks don’t have any idea how networking works, but that also means that it’s an area ripe for innovation, especially in the complexity-reduction/OpEx sense. NSX and open network switches, anyone?
I may seem negative about EVO:RAIL, but I’m actually quite positive about it. It is a 1.0 product, destined for future improvements, and a software vendor has to ship sometime. I also like the idea of additional competition in this space. There are so few real competitors that adding another will be good for consumers. Competition drives up innovation and drives down costs. Further, we haven’t seen pricing for EVO:RAIL yet, so there could be a major value proposition here as well. Finally, I expect that a number of improvements, such as the HTML5 interface, will find their way into the upstream vSphere and VSAN software. Everybody will benefit from that.