Back in the 1960’s, if you called music, movies or artwork “heavy,” well, that meant they were cool. Today the term takes on a whole different connotation. The heavy file sizes of video and audio, as well as stock photography, E-learning environments, telemedicine and many other applications, have made streaming media (along with the need for wider, broadband pipes or longer downloads) a practical reality. An obvious question is “Where does all this media get stored?” At least the question is obvious to Isilon Systems.
This enterprising, 18-month-old Seattle company has just come out of stealth mode to announce how its patented hardware and software solution will help companies deal with these large multimedia data requirements. By breaking out of the box, and into a cluster of smaller, interleaved, appliance-like servers, Isilon Systems is providing a new distribution source for information that taps the power of grid computing on a miniature scale. In essence, Isilon Systems founders Sujal Patel and Paul Mikesell have not only seen the future of managing rich media on the Internet, they’ve created the means to store and manage rich content more efficiently and cost-effectively than ever before. Investors have taken note. Isilon’s $8.4M first round of funding, led by Atlas Venture, marks a reawakening in the high-tech Internet sector.
We met up with company co-founder Sujal Patel to arrange our own data transfer of the thinking behind Isilon’s new product plans. Prior to Isilon, Sujal spent nearly five years at streaming media pioneer RealNetworks where he was responsible for their second generation architecture, networking and their distributed stream delivery technologies. Here’s the data download on what we learned.
Seattle24x7: Isilon Systems has a timely role to play in the evolution of the Net – the storage and management of rich media content. Where did your time line begin?
Patel: When I began my career at RealNetworks one of the things that I saw was a huge growth in streaming media from its infancy back in 1996 to an astounding growth rate over the last six or seven year period. I realized that storage would become a major issue as streaming media customers began to scale their infrastructure.
Seattle24x7: In multimedia and rich media content especially?
Patel: In the Library of Congress there are 20 terabytes of books, that’s 20 trillion bytes of information, roughly 20,000 copies of the Encyclopedia of Britannica. Now that we’re in the new millennium, the new measure is the rich content stored there. There are 500,000 movies, there are 13 million photographs, and there are 4 million geographical maps. This rich content is over a thousand times larger than the books themselves.
Seattle24x7: Industry pundits have also talked about evolutionary increases in computing speed and scale, correct?
Patel: Moore’s Law states that CPU cycles double every 18 months and that means that the processor that goes into your PC or goes into an enterprise support system is twice as fast as the one that went into it 18 months ago. That trend alone has left us with over a 50X increase over the last decade.
Bandwidth increases are even greater than Moore’s Law. George Gilder has written that “Bandwidth grows three times faster than computing power.” And so if you look at the world today, we’re in a world where CPU power is abundant and bandwidth is abundant. The third leg of this is that disk packaging costs have decreased as disk sizes have increased. In the last five years we’ve seen disk costs per gigabyte plummet over 100 times. If you go to Fry’s today you can pick up a 60 gigabyte disk and it costs you next to nothing. But what you can also do is you can walk down that same aisle at Fry’s and pick up a disk that can store every song in MP3 format that was made over the last hundred years. Disk capacity has improved dramatically.
Seattle24x7: The cost of storing data represents a lot more than just media, bandwidth and drive space?
Patel: Storage is really all about managing information. People want fast, secure reliable access to information and they want it from any device. They want it from their PC, they want it from their handheld. And right now, managing content, especially the rapidly growing data stores of rich media, is one of the biggest nightmares for customers.
If a business has a terabyte of data in its enterprise today, they’re usually spending tens to hundreds of thousands of dollars a year just to manage that terabyte — in backup and support, and the management of growth within that support.
Storage is also incredibly difficult to manage. The fact is that the traditional storage systems that were developed over the last decade aren’t scaling well as we grow rich content. The characteristics of rich content are different, the size of the data storage and the way you deal with rich content media is very different.
Seattle24x7: How is Isilon addressing these issues?
Patel: Isilon is creating intelligent software that not only reduces the cost of buying storage but dramatically reduces the time associated with managing it. And it frees administrators in corporations to do more important things than babysit storage.
We’re making storage easier to manage, we’re making it so that storage can heal itself, and we’re greatly increasing access to information. After all, that’s really what drives the information age. It’s really this access to information. That’s the future of the Internet business as well.
Seattle24x7: How does the technology work?
Patel: Essentially, the file system allows us to pool together different devices and form one unit that customers can manage from a single point, and that customers can view from a single network drive. In addition, we’re allowing the administrators to grow that seamlessly just by adding new devices to this cluster. Our technology has the ability to withstand the failure of any of the boxes in the system without data loss. You no longer have to contend with refrigerator size systems or very highly fault tolerant systems because you’ve got a cluster system that together forms one reliable unit.
Seattle24x7: The cluster is able to heal itself?
Patel: We have the ability to spread error correction throughout the cluster so that when any member of the cluster goes down, we’ve got the ability to reconstruct that data on the fly. It pushes the concept of RAID which protects against disk failures all the way to the physical computer level.
Seattle24x7: In a sense, you are deploying the power of grid computing?
Patel: Grid computing allows you to distribute computing jobs and databases across numbers of servers using the network to share resources and pool storage together. Isilon is applying this same concept to manage storage within an organization. We have taken advantage of all of the advances in technology and created an architecture that maximizes disk capacity, bandwidth, and processing speed. We pool this together and allow a customer to manage huge stores of data all from a single point. In the future, we will be able to apply this concept across geographically distributed locations.
Seattle24x7: You have spoken out on the public policy issues surrounding media content and copyright issues. How do you see the debate playing out?
Patel: There are policy issues that need to be addressed for technologies like ours to take off. Napster was one of the first steps toward grid computing, and towards distributed computing, and we’re all aware of what happened to Napster. There’s really a chicken-and-egg problem that you have to address. And I believe that problem starts with content.
Seattle24x7: How so?
Patel: Content is one of the most important drivers of broadband usage. And companies like Isilon will depend on the continued growth of content, be it digital entertainment or any other content-driven applications. We’re also banking on the fact that there will be an extensive broadband network within the U.S. to make this a reality.
When you look at the state of broadband today, nearly 70% of US home have access to broadband connections. But only one in ten of those homes actually subscribes to high-speed Internet services. At the height of its popularity, Napster accounted for over a quarter of the data in and out of residential customers, from just a single broadband service provider’s network.
For businesses to continue to provide broadband services, copyright owners must work with the technology community to come up with compromised solutions that will be mutually beneficial and of interest to consumers. There are a lot of technologies, like a digital watermark, that can help protect the content that these intellectual rights holders have. But the music and film industry also needs to make changes in their business model that will allow them to adapt to the state of digital content today, just like the movie industry and the television industry did when VCR’s first came on the market a couple of decades ago.
Seattle24x7: Thanks, Sujal. We look forward to watching Isilon’s future moves in helping Internet companies develop and deploy their information assets — on a major scale.
Larry Sivitz is the Managing Editor of Seattle24x7.