It is a great solution for unstructured file data and it works pretty much at any scale. We focus on the high end of the market because the solution has so many benefits when you have scale, when you’re trying to store hundreds of terabytes, when you’re trying to deliver those files to 20 to 30 locations around the world. That’s when the solution is really valuable.
If you’re trying to run an ERP system with databases, that is high-performance block data. There are many great storage companies that do an awesome job at high-performance block data, and there are application vendors that do great block synchronous replication, typically across two locations for active/passive DR [disaster recovery] of databases. That is not what Nasuni does. If you want to protect an Oracle system, if you want to protect your SAP system, that’s not what Nasuni does. That is the block database world. That is the traditional SAN world.
We are the NAS world. If you have files at scale, it’s a general purpose file system. I love calling it that because that’s the only useful kind of file system. We actually have many deployments where clients will upgrade to the latest flash array SAN storage product, something like Pure or Nutanix, and immediately have great performance. But, they say: What do we do with all the files? They’ll get a virtual machine inside these high-performance, solid state flash arrays and put their files there and their files will drain right out the back into the cloud. Now you have a really high-performance environment for your files, but you don’t have the complete file footprint in your data center. That kind of fine optimization can be extremely powerful.
It’s not unlike the world of thin laptops or smartphones in that everything has gone to solid state. Everything is super high performance on the local handheld device that you’re carrying. The thing we’ve done with those devices though is that we squeeze the capacity out of them. There is much less capacity in those devices than we used to carry around five or 10 years ago because the data is really in the network. Whether it’s your email, your music collection, your photo collection, the smartest thing that Apple ever did was put all the consumer data in the cloud so that people could get thinner, faster laptops at an affordable price.
When I want to move from one generation of Nasuni hardware or Nasuni VMs to the next generation of my SAN hardware infrastructure, there is no forklift upgrade. It’s no longer: Oh my God, I have to take 100TB from this monolithic NAS storage controller to this other one and it’s going to take months to do that. The upgrades look a lot more like getting a new iPhone. You get the new hardware, you put it online, you tell it this is my new NAS device. This is my new Nasuni edge appliance. You connect it to the service and it resynchronizes with everything that was there in the old appliance and nothing came from the old appliance. Everything came back from the cloud. It just streamed back from the cloud. It’s safe and it’s much more convenient than doing this bulk migration every three or four years.
Do you tell customers to replace your existing storage with Nasuni, augment it with Nasuni? What is the strategy?
We tell them we will change their business. File systems are strategic. File systems change how people work in organizations. They are the ultimate duct tape for an organization. You’re going to have your document management systems to create some workflows and yes, you’re going to have your ERP systems for supply chain management. But when you’re talking about groups of people getting work done, it’s about the files: Who has access to the files, who can change the files and who can see the files around the world? There has been a bottleneck in the industry around file systems for at least the last decade. We’re going to liberate you from the bottleneck around your files. We can get great collaboration because the performance is really terrible across all these sites. We can get rid of that. We can give you high performance on every single site.
This is, I think, one of the more interesting things. We are seeing the tip of the iceberg around machine-generated data in the enterprise. We have several large manufacturing companies that have been constrained in terms of how much data they can store and analyze, by how much data they can store financially and technically in their exiting file systems. All of a sudden what you’re seeing is the same kind of scale that hit the web companies 10 to 15 years ago that caused things like S3 to develop, that caused things like Google to scale their infrastructure.
Those are now hitting the larger segment of the enterprise where you will have a big manufacturer saying: Wouldn’t it be amazing if we could have telemetry on every single device we have out there? Where are we going to store it? We’re not going to want to access it from one point; we’re going to want to access it from everywhere, anyone that wants to consume this service to make our products better. Or we want the test data for how we manufacture things to be stored forever so that when something starts failing in the field we can determine where we missed things and what happened. Nasuni will get rid of all your previous problems around files but, more important, we give you a file system strategy for the future that is going to make your business far more productive, far more responsive to your own customers by helping you build better stuff.
Are your customers taking out existing storage or just not buying additional future storage?
There is a combination. Many customers have taken out their storage and their backup systems. That’s kind of a bread-and-butter thing that we do. It could be an EMC-native environment and they’ll have backup environments around that. We can eliminate all of that and you can store with our edge appliance and Nasuni service and you’re done. You don’t have to worry about making backups anymore and you don’t have to worry about ever running out of space. That’s one side of the equation. The other side is global access to the data. That is brand new. Customers have been stuck for years or decades with essentially [this problem]: If you have a lot of files in one location you have to go over the wide-area network to get those files with some kind of network acceleration product.
That approach is super frustrating to companies. When you want to have 40 locations around the world doing software development/testing and have them all feel like they’re operating on a local file system, the approach should be that the file sits in the cloud but it’s replicated locally to every site where that file system needs to be in use. That radically changes what could be done in the past. It creates new workflows that for this client are one of the most exciting things they’ve seen in their business in a long time in terms of being able to get more work done faster.
Just a very quick tactical question. What happens if your site is offline and can’t get to the cloud?
If you are not trying to do one of these global synchronization things the system behaves like any kind of local storage system. The same with your Nest thermostat; when it’s offline it’s still a good thermostat that you can control and you can make rooms warmer or cooler. We fall back to the behavior of essentially being like a traditional NAS box. Everyone can keep writing and reading from it. The cache helps you a ton because typically everything that people need is right there and when the internet comes back online the device is synchronized to the back end. They are all designed to work offline.
I’ve been talking about one file system but our customers typically have dozens of file systems, not because any one file system couldn’t scale forever but because they want different behaviors for their file systems. They may want to have a file system that is local to a certain site and another file system that is local to a region or that is global. You can change the behavior at the file-system level. If you have this kind of global access and the internet goes down, you can tell the file system at that site to exist in read-only mode for the data that they already have so they cannot make modifications and those modifications don’t conflict with what people are doing elsewhere in the world. Or I want to have a site fall into read/write mode and then our system will identify conflicts. It never throws away any data and says: Your end users need to now reconcile these files because you’ve allowed conflicts to happen in the system.
Why couldn’t I do what you’re doing today with my existing file systems and cloud storage? Why couldn’t I create that same hybrid cloud approach with what I already have?
Because it’s all about the file system. To solve your file problem, you need an awesome file system. We have built the first cloud-native file system. Before I started Nasuni I started an object storage company so I knew how cloud storage systems had to be built. My previous experience was all in distributed systems and I knew we needed a file system that could be replicated and live comfortably among tens of thousands of servers distributed all over the world. That’s what UniFS is. It’s a file system for the times. It’s a file system that, at a technical level, has a limitless pool of metadata to draw from in terms of the cloud and therefore it can scale forever.
You can take a UniFS file system and go from a million files to a billion files and not change the device. That appliance in the front doesn’t change at all. All that file system growth happens in the cloud and it scales forever. We use that same technique of scale in the cloud to scale across time. The reason we can store versions forever is that the file system is able to map in time the same thing that it’s doing in space with all the metadata and basically give you full-on versions of your files every five minutes for an infinite amount of time. There is no file system that can do that in the cloud providers, in the traditional vendors. That’s what we do.
Finally, providing access to that file system from multiple locations is not just about having one file system that can live centrally - which we have because our file system is in the cloud - it’s about the orchestration and management of all those edge appliances around that file system core. It’s really important to be able to maneuver the blocks and the state of the different synchronizations that are happening with the edge appliances from each location to that file system core. That is what Nasuni back-end services support. They’re orchestration wrappers around UniFS to make sure we are not stepping on our own toes as we’re trying to synchronize 30 locations around the world on read/write, active/active endpoints.
Nasuni vs. legacy players
I know that there are some specific competitors that come up in relation to your company but I want to talk about the traditional storage suppliers at this point. Is there any indication that they’re doing something or are planning on doing something similar?
The traditional competitors are NetApp and EMC. Those are the ones that we see all the time. EMC has VNX, which is their mid-market NAS array, and they have Isilon, which is what they use for scale. NetApp has NetApp. They have cluster and they have their WAFL FS. What’s limiting about their approach is they are stuck in the mode of thinking everything has to live in our devices. Everything has to be in our arrays.
Recently, what they’ve finally said is: We are going to keep the file system in the arrays and all the metadata for the file systems in the arrays, but we’re going to allow you to tier to the object store. I use object store and cloud storage as synonyms. We’re going to use the cloud storage systems to tier. This is typical of dominant players. They see this new technology as a second-class citizen and relegate it as a second-class citizen in their architectures.
If you have an Isilon array and they allow you to tier to the cloud, they are essentially keeping all of that file system metadata in the Isilon array and then there are pointers back to the cloud. The problem is that to get a bigger file system you basically need to scale the front end of the cluster. And to be able to distribute the cluster around the world it’s the same old approach. You need to go back to that cluster because the file system is located on that physical cluster that you still have in your data center. The way to think about this is to completely turn the file system upside down and say: No! The cloud storage system is the first-class citizen here. The file system needs to exist there and then you need to have this disposable, stateless device that extends it.
That’s the beauty for consumers. That’s the Dropbox approach. The files you can list with the service and that gets extended to wherever you want to use it. That is a much better experience than trying to synchronize all your devices against each other. Traditional vendors are still in that world.
What do they tell customers about you when they’re selling? What do they warn them about?
They are trying to sell them a thermostat with tons and tons of buttons and whistles and they tell them to look at all the buttons, look at all the configurations that you can have with our stuff because we’ve been doing data center devices for the last 20 years. Nasuni is this streamlined, very simple appliance that cannot possibly be complete. Every once in a while there will be some bell or whistle that we don’t have and we have to put it in the road map.