Converged infrastructure comes in many forms. Some vendors put a bunch of discrete hardware together on a pallet and call it “converged.” Others think that their use of iSCSI or FCoE means they’ve got converged storage. Yet the real holy trinity of convergence is when a vendor converges compute and storage resources on three fronts: acquisition, implementation, and ongoing management. This is where Nutanix operates.
Nutanix is both a hardware and software vendor. On the hardware you purchase 2U rack-mountable building “blocks” and fill them with up to four nodes each. There are two models of NX-3000 hardware, the NX-3000 itself and the NX-3050 with a different mix of CPU and storage. Regardless of the model, each node has two eight-core Intel E5 2600-series CPUs and up to 256 GB of RAM. Every node also has 400 GB of Fusion-io PCIe flash memory, 300 GB of commodity SATA SSD, and five 1 TB commodity SATA drives spinning at 7200 RPM to form three distinct tiers of storage. Linking each node to the others is dual 10 Gbps and dual 1 Gbps Ethernet. Nothing in this list is particularly special, given the state of blade computing, yet Nutanix turns it into something distinct with their software.
On the storage front, the NX-3000 doesn’t use traditional storage, though there’s nothing stopping you from connecting to your existing IP-based arrays. The whole point of these devices is really the Nutanix Distributed Filesystem (NDFS), designed by some of the same folks that built the massive distributed filesystems that underpin all of Google’s operations. NDFS uses all the flash, SSD, and SATA disk on each node to provide resiliency and fault tolerance, automatic tiering with data compression, and snapshots, backups, and remote replication functions. In fact, with their latest software releases they have the ability to do multi-master replication, a feature that has traditionally only been available on expensive EMC VPLEX hardware as well as midrange Dell Compellent storage arrays. This allows for a variety of stretched data center and business continuity scenarios that aren’t typically possible with the active/passive nature of most storage replication. It’s likely there are some caveats, usually having to do with distance and latency in these cases, and you still have to figure out the networking, but this is really an example of how the software defined data center can excel in certain ways.
Nutanix delivers the NX-3000 preloaded with either VMware vSphere or KVM, for OpenStack compatibility. They also have a single-pane management interface that allows visibility into performance and capacity on a per-VM level, which is often hard to do with traditional storage. As a comparison, Tintri’s VM-aware storage is likely the closest thing to this while trying to achieve similar levels of simplicity. Adding to or upgrading your Nutanix-based virtualization clusters becomes a task of buying the new hardware, cabling the networking and power, and telling the management interface to go ahead and use the new equipment that it has auto-discovered. Pretty simple. They also offer a command-line interface for those that feel more comfortable or would like to automate even further.
So should you buy into a system like this? In many ways the Nutanix NX-3000 seems too good to be true. To a large extent it is a black box, which is scary to many traditional IT people who expect hundreds of knobs and levers to turn and tweak. There are some knobs, like telling the storage how many replica copies to keep, but much of the traditional complexity is just plain missing. IT staff have better things to work on, but getting people past the mental hurdle is tough sometimes. I, too, often wonder what I would do with my day if I didn’t need to chase down obscure performance problems or do a lot of unnecessarily complex provisioning.
One big thing when considering the NX-3000 is that you have to be committed to virtualization. Unlike other converged solutions like the VCE Vblock or the NetApp FlexPod there’s no room here for bare-metal use of the equipment. The nodes aren’t particularly large, either, which may present a problem if your workloads don’t scale horizontally. But that might be just fine. Consider the case of virtual desktops, an area where Nutanix has been stepping up with newly-written VMware Horizon reference architectures. Each node is easily deployed, a fixed cost, and a known performance quantity on all fronts, all desirable traits for VDI deployments.
If you look at where data center technologies are heading, we see tiering using flash and SSD, we see blades, and we see distributed, software-defined storage, like the vSAN functionality VMware has shown tech previews of at VMworld 2012. Thing is, these are all things that Nutanix has right now. Pretty interesting, if you ask me.