M&E Daily

Qumulo: Pushing the Boundaries of What Storage Can Do

By Chris Tribbey

Seattle-based Qumulo is all about improving how companies deal with data storage.

The co-founders of the three-year-old software storage company were the brains behind Isilon’s award-winning OneFS distributed file system, and between them they’re named on 55 of the 62 patent inventorships for Isilon. In two rounds of fundraising, Qumulo has already secured $67 million from top venture capital firms, and currently has more than 15 petabytes deployed at customer sites via its scale-out network-attached storage (NAS) solution.

Brett Goodwin, VP of marketing for the company, sat down with the Media & Entertainment Services Alliance (MESA) to discuss the challenges in handling large-scale unstructured data in the media and entertainment space, the importance of metadata and analytics, and what’s changed since the first iteration scale-out storage solutions.

MESA: How did Qumulo first come about … what gap in the data storage market was the company looking to fill?

Goodwin: When we founded the company we took a very research-driven approach, and over the first two years of the company, we did over 600 research interviews, with storage administrators, storage buyers, storage architects, to really understand what was working well today, and what their real pain points were. We found things that surprised us: unstructured data is growing at an incredible clip, and, in general, the amount of unstructured data for companies we’re talking to doubles every few years, or more quickly than that. Finding an economic solution that allows them to scale out their available storage, without adding additional labor to manage that storage, and to do it in a way that fits into their budget, is a real challenge. That’s one of the things we set out to do.

The more surprising thing we found, was that the nature of the problem has changed: where it used to be ‘How do I manage my storage?’ the challenge today has become ‘How do I understand and manage my data?’ I now have millions, billions, on my way to trillions of files and digital assets. How do you understand and manage this growth and how fast it’s growing, who’s creating it, which users and applications are accessing it, what’s consuming performance on my system, and what files haven’t been touched in a long time? So being able to answer questions about your data at scale is something storage has not been able to do in the past…

Qumulo has introduced the world’s first data-aware, scale-out, mass network-attached storage, and it’s file- and object-storage for high-performance computing and extremely large-scale unstructured data. Our goal is to make storage invisible, and your data instantly visible. We’re a software-based solution that runs on top of Linux, and runs on 100% commodity hardware. Unlike systems from the past, which typically were tied to proprietary hardware, we are a software-based solution that runs 100% in the user mode on Linux, and is designed for commodity hardware, and the commodity hardware we run on uses a mix of Flash and high-density spinning disc, an optimal balance of performance per dollar, and capacity per dollar.

MESA: Specifically in the media and entertainment space, how does Qumulo’s network-attached storage (NAS) offering help M&E companies get the most out of their assets?

Goodwin: Three of the top five of the world’s biggest animation studios use our product, for large-scale rendering of state of the art animated motion pictures. They have a need for extreme scalability performance, and parallelism, so they can render all of these frames and all of the motion picture in the time that they need to in order to hit their release date. Another customer of ours is Sportvision, which does the yellow first down lines you see when you watch football, and other enhancements to the sports viewing experience. Another customer of ours does CGI and special effects for games and motion pictures.

The commonality here is they have very large, ever-expanding data sets. As we move from standard definition, to high-definition, on our way to 4K and beyond that, the file sizes, the frame rates, the sheer volume of file and asset size continues to expand at a break-neck pace. They need a system that they can linearly scale with capacity and performance in lockstep. With the Qumulo system, you start with a minimum of a four-node cluster, a scale-out system where as you need to add performance and capacity, you simply add additional nodes. Sportvision started with four, went to eight. We have customers running 20-node clusters, with over four petabytes in a single cluster. This gives them this modular flexibility to pay as you grow.

Media and entertainment companies using us also include post-production house FotoKem, the video production studio Blind, which makes everything from Audi commercials to Coldplay videos, as well as Zoic Studios.

MESA: How do metadata and analytics come into play with Qumulo’s media and entertainment offerings?

Goodwin: We build extended metadata and real-time aggregates right into the file system itself. One of the things we’ve demonstrated at the National Association of Broadcasters Show and SIGGRAPH (Special Interest Group on GRAPHics and Interactive Techniques) was a single file system that had over 10 billion files on it, over a million directories, and you could get real-time information about your system.

At a single glance, from a single dashboard, you could see how much of your storage was being consumed by directories and sub-directories, what the read and write performance was of the overall system, as well as mapping that down to the different projects, and who’s accessing what files and applications, in real time.

This kind of information hasn’t been available to people responsible for managing the storage and infrastructure.

MESA: How does your software solution, Qumulo Core, stand out as a next-generation scale-out solution?

Goodwin: Real-time analytics are built directly into a built-from-the-ground-up distributed file system, which we call QSFS (Qumulo Scalable File System). It’s a software only solution, that runs on a standard OS, designed to run on commodity hardware, so you can run this on appliances that you get from Qumulo, or for large-scale deployments, we can certify other hardware platforms. ..It also can run as a cluster of virtual machines, so if you wanted to test it, I could send you a download, and you could build a cluster of four virtual machines, which makes us cloud ready. On our roadmap is extending this capability so it can run in the cloud, or on premise…

We’ve made our system 100% programmable, so everything you can do through the Web user interface — simple, elegant, easy to use — you can also do through an application programming interface (API). You can easily program the system to do automated tasks, whether it’s provisioning storage or accessing the metadata analytics, everything is accessible via the rest API.

MESA: What about security, with so much data and so many moving parts?

Goodwin: We integrate into companies’ active directory structures and existing security and permission systems. In general, our system is deployed in their own data center, behind their existing firewalls and intrusion detection systems. Down the road on our road map are things like encryption of data. Today, fitting in the existing system frameworks that media and entertainment companies have is our approach when it comes to security.

MESA: What’s ahead for Qumulo? What hurdles in the enterprise data storage business would the company like to go after next?

Goodwin: Our mission is to become the company that the world trusts to store, manage and curate its data, forever. And we feel like there’s an opportunity out there to become the next major player in the storage industry. And we love the media and entertainment industry because it’s so forward thinking, always pushing the boundaries, coming up with new ways to move the user experience, and push what today’s storage systems can do.