When I was attending Interop in Tokyo a couple of weeks back, I had a few participants ask me the same question: how is Nutanix different from other “converged appliance” vendors? At first I was a bit flummoxed by this question because I didn’t know any other vendor that did what we do. But when I listened to them talk about it, I realized that with all the marketing machines going full force, a decision maker in a company looking at new data center infrastructure has every right to be confused. The cause for this confusion is stemming mainly from two sides.
1. Big vendor’s stack approach:
The established players have recognized, rightly so, that customers are thirsty for complete and converged infrastructure for virtualization projects. They went about it the only way they know — by creating alliances among equals. To me, it’s like tasting soup where all the ingredients don’t really belong together, but just float around in the broth and you have to eat it since you are hungry and you paid for it. NetApp’s Flexpod and EMC’s Vblock come to mind. The compute, storage, networking, hypervisor, fabric switches, and SAN management software are all packed in a rack, delivered with the promise of integrated customer support and a unified user interface. You know you have a problem when the datasheet for the product is a one page and the actual quote is five or six pages.
But then again, that is all you can really expect from legacy vendors with hundreds of thousands of existing installations.
2. Startup’s “whatever-we-have-is-converged” approach:
Some startup companies realized that customers are looking for converged platforms. They just hitched their wagon to the train and called whatever they are doing “converged.” Take, for example, one company with a page titled “Introducing converged storage” on their website. How is their platform converged? According to them, because they can keep their primary and backup on the same device. Let’s ignore for a moment if that is even a good idea. Let’s look at what that means behind the beautifully crafted marketing theme. Take some flash/SSD, take a few hard drives and string them behind a Super Micro server, use some variant of FFS or ZFS that provides versioning and compression (not really de-dupe, right?) and say, here is a solution where you can keep your primary and secondary data. The question remains, however, what happens if the flash/SSD fails? Having another x86 server standing doing nothing really isn’t clustering; it’s just a cold-standby vintage 1991.
Convergence is not about increasing the risks; it’s about eliminating or at least reducing the risks within infrastructure layer.
Yet another startup is jumping on the convergence messaging bandwagon by saying it’s about converging management of compute and storage. They claim that by allowing the storage to be managed and monitored from VMware vCenter, they solved the convergence puzzle. Never mind that your compute servers are running separately from the storage servers (which are pretty much identical hardware) over the network via NFS. Yes, they removed LUNacy alright, just to replace it with smoke-and-mirrors.
Convergence is also not about scratching the surface on management complexities and operational expenses.
While I was in Tokyo it was raining heavily and one day I happened to see a vendor outside the Ginza train station selling umbrellas for 300 Yen (around $4). These were not the best quality, but he was doing brisk business. He was at the right place at the right time.
When flash and SSD technology took off, there was a real vacuum in the channel left behind by the acquisitions of EqualLogic, Isilon, and Compellent. A few companies just happened to be there to take that business off the table. Right time, right place. Not that there is anything wrong with it and I wish them luck! But I personally believe that startups are less about marketing gimmicks and brute force sales and more about strong technological innovations.
Visionary IT directors and architects don’t get away from the comfort of their legacy vendors and come to an emerging technology provider because they want timid, incremental innovation. Given enough time, they can get that from big vendors. They come to startups because they want bold, market-making solutions way before their competitors can, so that their IT infrastructure becomes a key differentiator and a crucial competitive advantage. In the mid 1990’s, NetApp did it for its customers; in mid-2000’s, Data Domain and Riverbed provided such path-breaking technology to their customers as well.
This is exactly what Nutanix has set out to do. We didn’t set out to build a cheaper faster mousetrap. We didn’t just build a single (or dual) controller storage system using a standard x86 server — that is what vendors did in the 90’s. We built a truly converged platform that eliminates the need for you to buy any more legacy servers or legacy storage for your datacenters. We recognized the opportunity in convergence before anyone else did, but we also recognized the key challenge that comes with it. That is, when you converge, you shrink the available resources. You are now doing more with less, which means you need to have an effective way to scale the environment when you want to. Hence, clustering is paramount. Nutanix addressed the need for clustering in three ways.
First, Nutanix brought Google-like, cloud-scale clustering into the realm of enterprise data center by integrating it inside the VMware hypervisor. This clustering is unlike anything else because of how efficiently it scales. You can read about it in other articles in our blog. Scaling linearly with a factor of one is not an easy problem to solve. (If it were, the aforementioned “converged” players would have done it.) We let our customers start small and scale to thousands of nodes at their own pace.
Second, we changed the performance (IOPS and latency) game by eliminating the need for any type of network between your applications and your data. (No iSCSI and no network-bound NFS.) But we just didn’t stop there. We brought the best-in-class PCIe flash—Fusion-io—directly inside the VMware server where your applications are running and made it a primary storage device for read, write, cache, data management, and metadata. Check out how EMC has been trying to do this incrementally with Project Thunder and VFCache initiatives.
Third, while we stand behind the goodness of our NoSAN message, we do acknowledge that SANs provide some very important data management features such as snapshots, cloning, thin-provisioning, and replication. We couldn’t ask customers to leave all these features behind, so we built them natively into our platform and tied them, where applicable, into the VMware’s vCenter management APIs.
There are times when you have to take a scalpel to a problem. But in the technology sector, once in a decade you get a window of an opportunity to take an axe to a very complex set of problems. The emergence of flash as viable storage media, the prevalence of hypervisors as the delivery vehicle for almost all applications, and the growth curve of the x86 architecture have provided that unique opportunity a lot of IT architects have been waiting for. At this time, Nutanix is the only provider with a solution that takes advantages of these conditions.
Nutanix is not here to make a quick buck on a rainy day — not that there is anything wrong with that. We believe that, if executed well, Nutanix has the potential to change how data centers are built in the future. We and our customers are having a lot of fun building these next-generation data centers, and in the process, building a company that will make a difference.