The Network is the Computer

Allow me to once again apologize in advance for a lack of brevity. This is one of those blogs you wait a career to write.

A few years ago, I was sitting across from a Wall Street CIO, one of many I was visiting in New York. I was asking them all the same question, "do you feel the grid you're building is delivering a competitive advantage to your business?" (For those that don't know what a grid is, it's a collection of low cost network, storage, computing and software elements, lashed together to do work that historically required very expensive dedicated proprietary technologies). I asked the same question of CIO's in the energy industry, using grids to find oil. In the life sciences industry, using grids to discover drugs or model proteins. In the movie industry, using grids to render movies.

The answers I received, typically delivered by an impassioned CTO that had spent a year building a grid, was always the same: "absolutely yes. Our grid is way better than any of our competitors'."

I haven't stopped asking that question. But about a year ago, after Sun outlined plans to build a public, multi-tenant grid (just like the power companies run), and make it available for $1/cpu-hr, and after a few industry notables began suggesting change was afoot, I started hearing a different tune. "Um... maybe my grid's no different than anyone else's."

Now, since John Gage first uttered the phrase, Sun has been saying "The Network is the Computer." It's one of those rare vision statements that only becomes more true over time. And next week, we're going to prove the point by unveiling the world's first on demand supercomputer. And by on demand, I mean accessible through your browser, with a credit card. This isn't yesterday's definition of On Demand, involving custom financing contracts, prepositioned inventory and a sales rep in a crisp blue suit ready to negotiate. Nope, our definition is just like eBay's: you bring a browser and a credit card, we offer the service. No fuss, no muss. We believe the simplicity, accessibility and affordability of this service changes the face of computing for all organizations, large and small, public or private.

The Sun Grid (which will be officially unveiled in a few days) is an offering we and our partners will be expanding over the months and years to come - like any good product, there's no end to the innovation possible. This represents not only the future of product development at Sun, but like the Java platform and the internet itself, it really represents the future of computing.

As strange as it may sound, consumers are way ahead of most enterprises when it comes to using grids (and paying for them). Most of us live on the grid at home - we use Google and Yahoo!, we love eBay, we upload and share photos and movies, and gather our news from various sources on the web. Most of us bank from home, we leverage network email services - and if you think about it, that transformation all occurred within the last decade. In the blink of an eye.

But behind the corporate firewall, the transformation toward multi-tenant grids has been slower. Frankly, it's been tough to convince the largest enterprises that a public grid represents an attractive future. Just as I'm sure George Westinghouse was confounded by the Chief Electricity Officers of the time that resisted buying power from a grid, rather than building their own internal utilities. But that's not to suggest it hasn't been happening in the business world.

Witness the meteoric rise of Salesforce.com - or RightNow, or PayPal - or any of a number of other services designed to replace traditional infrastructure behind the corporate firewall. Smaller businesses especially have flocked to the grid to spare themselves the headaches of architecting and owning their own datacenters.

But larger enterprises have been tougher to convince. As an example, for the past 15 months, we've been negotiating with one financial institution interested in leveraging our grid for spike loads of portfolio simulations. When their procurement team held up the contract to start negotiating the gauge of chain link we'd use around the grid, and which vendors were approved to supply network cables, we gingerly passed them back to our traditional sales channels - this was clearly a customer that would prefer to build their own infrastructure (can you imagine arguing with PayPal over chain link?). So be it, that's where most IT is purchased today, and will likely be purchased for decades to come.

But there's no denying there's a change occurring.

A good friend of mine, a bioinformatician (love that title), once described how frustrated he was at having to wait for his university's supercomputing facility. "If you had a grid available on line, I'd bring my whole budget to you." Granted his budget was something like $10,000 a quarter, but rumor has it there's a good business in the long tail. My view - most computing will be purchased by that tail. There are, after all, far more small financial institutions than large. The same applies to movie studios, pharmaceutical companies, academic institutions, and nearly every other industry on earth. I'm very comfortable betting on the value in volume - and the willingness of those smaller firms to change culture, process and lifestyle to get a competitive advantage through network services. Just think back ten years - when most enterprises I met laughed at the idea of putting business systems on the internet. Now you're an anomaly if you're "off the grid."

But getting to this week hasn't been without hiccups. After we announced it, we started working with a number of companies interested in negotiating the equivalent of chain link fencing, as above - we saw IBM Global Services (and HP's equivalent) in every one of the deals. We learned a lot, but mainly that most enterprises today define On Demand computing as hosting - they want to give their computers, software, networking and storage to a third party, and rent them back for a fixed price. But that'd be like an electricity company collecting generators and unique power requirements, and trying to build a grid out of them. That's not a business we're in (nor one in which technology plays much of a role - it's all about managing real estate and call centers, as far as we can tell). Grids are all about standardization and transparency - and building economies of scale.

Building a secure, publicly available multi-tenant grid also turned out to be exceptionally complex - there's a reason no one had ever done it before. Most grids are application specific - for search, or auctions or payment. A general purpose computing grid was ploughing new ground - and we wanted to ensure availability and security would be as high as possible. To stress the grid, I actually sent mail to all of Sun's employees challenging them (with the promise of a new workstation) to see if they could bring it down. On the theory I'd rather have a Sun employee, especially a Sun engineer with deep insight into our products, show us how to break it, than a rogue user.

After disappointing a huge swath of our employees who couldn't participate in the contest (our export control policies constrain which elements of our global workforce can be exposed to the grid), we surfaced several vulnerabilities in the very high-scale interaction of hardware, networking and software platforms (again, given that no one's ever done this before, it wasn't all that surprising). We also engaged with the folks who monitor technology export control for the US Government (if there's a harder civil service job in the government, I'd like to know it) - who helped us ensure the grid wouldn't be accessible to people with nefarious intent. They understood we wanted to make this as simple as applying for an eBay account - we'll be close, but we've got to have a higher level of scrutiny (which is why, when you apply for an account, it'll take a few hours, and won't be instantaenous - but that's our goal).

Those are just a few of the hurdles we faced, but now we're ready - ready to release the first, publicly accessible instantiation of the future of computing. And here are a few things to be aware of:

First, in this first release, the Sun Grid will be available only to customers inside the US. Why? Export constraints. Stay tuned for international availability. And yes, we will be doing this globally.

Second, don't expect instant account provisioning. We're shooting for a few hours, depending upon demand, and no worse than 24 hours. But please be patient. We are focused on ease of provisioning, but we're also conscious of the risk and security requirements.

Third, we're opening on day 1 with less than 5,000 cpu sockets (both Opteron and UltraSPARC) - the world's most power efficient servers. As demand emerges, we'll be adding to that capacity - without limitation.

And finally, stay tuned for the web service API's. What you'll see this week is relatively simple, and a version 1.0 foundation for what's in store. Where are we headed? To release computing as a service, to be mashed up with other services (I can hear VC's around the world offering a standing ovation - 'no more having to build one datacenter per startup!')

If you're read this far, here's a final bit of color on the incredibly fortuitous domain name for the future of computing: Network.com.

As it turns out, midway through Sun's due dilligence in the acquisition of StorageTek, we learned they were the owners of Network.com. They hadn't really ever used it - a hidden gem. In hindsight, it may end up being one of the most valuable domain names in the history of computing. And we're certainly going to do what we can to burnish that value...

So have at it! Go to network.com later this week, grab a PayPal account, and experience for yourself what it's like to use one of the world's largest supercomputers. Without having to house it, manage it, power it, administer it, provision it... or buy it.

The Network is the Computer.

Once again, more true by the day.

Posted on 12:13AM Mar 20, 2006 | Comments[35]

Share this post  | | | | | |