Forecast: Cloudy

Who would have thought that infrastructure was the most interesting part of the Internet? Welcome to cloud computing, where big pipes and big iron create a second—no, third—generation of Internet entrepreneurs. Matt Cooney asks: could New Zealand become the Land of the Long Net Cloud? Plus cloud concepts and Me 2.0
Magazine layout

Koordinates founders Robert Coup and Ed Corkery: mapping the world from Auckland. Photograph by Phillip Simpson

Who would have thought that infrastructure was the most interesting part of the Internet? Welcome to cloud computing, where big pipes and big iron create a second—no, third—generation of Internet entrepreneurs. asks: could New Zealand become the land of the long net cloud?

Robert Coup and Ed Corkery are on a mission to bring order to this muddled planet.

Through their business,, the two Aucklanders are collecting as much mapping data as they can and publishing it online. If you need to know the position of every DOC hut in the Ruahines, fault lines in California or the exact slope of a hillside in Khandallah, Koordinates can tell you. That’s pretty handy information. If you’re a surveyor, an architect or a council planner, it’s gold.

Coup and Corkery collect map data from every source they can find, aiming to be “the one place for geodata”. Since Koordinates launched in April last year they’ve been adding data as quickly as they can find it.

Not surprisingly, all this data takes a lot of disk space and bandwidth—which should be a big problem. Coup and Corkery aren’t rich (yet), they’re not being funded by a kindly investor with deep pockets, and although they want Koordinates to be profitable, it’s early days. So how can they afford to make terabytes of data—by mid-year they expect to have ten terabytes online, or about 2,500 times the size of the entire Wikipedia—available to anyone who asks? Says Coup: “We looked at how we get up and running as a startup without spending an inordinate amount of money on infrastructure. How do we get it so that we can efficiently store this data, process it, access it, and make it available to people?”

Like most things, the answer can be found in a bookshop. The world’s biggest bookshop, in fact: In 2006, Amazon started offering space in its massive server farms to third-party developers. This service, called Amazon S3 (for Simple Storage System), allows anyone access to Amazon-grade web hosting, delivered over very fast connections to every corner of wired world.

The price of computing is going to drop considerably when an operating system and productivity software isn’t part of purchasing a computer … it will completely change the market. That’s when you’re going to see the real acceleration of cloud-based services, because everything will be on the web

It’s cheap, too: 15 US cents per month to store a gigabyte of data, and 17 cents to send that gigabyte to a customer. If you use more space and traffic, it gets cheaper. Many small S3 customers pay less than a dollar a month.

Have some number crunching to do? Amazon can help again: Elastic Compute Cloud (EC2) allows users to create any number of virtual computers, running Linux or Windows, and put them to use, all from your desktop. Pricing starts at ten US cents an hour.

This is commodity pricing, and S3, EC2 and their competitors—more on those later—are commodity services. They’re examples of cloud computing, a term that refers to computer resources that are deployed over the Internet and can easily scale up or down as needed. Koordinates’ developers don’t know where their servers are located, don’t know what hardware they’re running on, probably don’t even know when a server crashes. Amazon handles all the messy details; Coup and Corkery just know they have as much bandwidth, disk space and processing power as they need, and they’re not paying for anything extra.

The downside? Don’t expect too much flexibility: you can dump your data in S3 ‘buckets’, but you can’t then index them for searches or manipulate them as you could on your own disk. You can create virtual Windows, Linux or OpenSolaris servers in EC2, but don’t presume it’ll run that weird accounting package your cousin built before he realised he actually wanted to be a sculptor. And cloud servers can be as temporary as, well, clouds; Amazon warns developers that if an EC2 instance crashes, their data will be lost too. It’s up to developers to expect problems and work around them.

Still, cloud computing brings the capital investment for web startups to virtually zero, removes much of the workload of systems administration and allows companies to meet unexpected demand without buying any hardware of their own. And it also heralds some far-reaching changes that could revolutionise the way we gather, store, display and consume content of all types, from Koordinates’ map data to Hollywood movies.


The most competitive area of software development today is arguably in the web browser, and within the browser one area is the most competitive of all: JavaScript performance. JavaScript is the language that allows web developers to make web pages behave something like a desktop program. Apple, Google and Mozilla have all shipped much-improved JavaScript engines and announced even faster versions in their next releases, while Microsoft promises souped-up JavaScript in a future version of Internet Explorer.

Why? Surely the network is the main drag on performance, not the browser? Of course, the network is still a factor, but today most of the websites we use most often, from Trade Me to Twitter, are web applications. Many of the best rely on JavaScript to be as user-friendly as possible. Well-written sites also use JavaScript to fetch only the parts of a page that need changing, reducing the network load and reacting faster to user input.

These two factors—super-fast commodity services in the cloud and a rich, snappy experience in the browser—give software-as-a-service (SaaS) companies a ready advantage over traditional desktop vendors. There’s still a place for desktop software, but any company creating a software product from scratch now will ask if it can be delivered over the wire, and probably decide that it can. For every established, successful software vendor like MYOB, there’s a nimble SaaS startup like Xero that’s trying hard to eat its lunch.

Ask Xero’s CEO, Rod Drury, whether he plans to move his Windows-based infrastructure into the cloud, and he’s quick to say no. Xero, he points out, is not a typical Kiwi startup: it was funded from day one to employ 50 people for three years. By Xero standards, the servers, hosted at Rackspace in Texas, are not a big expense. “We spend probably about US$12,000 a month on all of our hosting—it’s about the same price as one-and-a-half people. The actual hosting and infrastructure cost is relatively small.”

It’s other essential things that cost the most, he says: marketing, customer service, feature improvements and user testing. “You won’t just turn stuff on in the US and suddenly you get thousands of customers. You actually have to do the marketing as well, because it’s very difficult to create noise. That’s the challenge!”

Drury is no cloud sceptic though. “I think the credit crunch means it’s going to be harder now for people to raise money. And there’s this massive cloud computing opportunity. We don't have the infrastructure to connect us to the world, and we don't have the investment culture. So at the moment I think there’s this huge opportunity to use brains—rather than producing stuff, real stuff with carbon and that sort of thing.  We’re really missing the opportunity.”

For Drury, a third factor comes into play: the desktop is losing its dominance. The iPhone is the first mobile phone that people actually enjoy using to surf the Internet (according to the AdMob Mobile Metrics Report, in February the iPhone generated half of all smartphone traffic in the US). Google’s Android promises to be a similarly web-savvy device. And the hottest computers this year are netbooks: tiny, lightweight, laptop computers, typically with a nine-inch screen, a reduced keyboard, no CD or floppy drive, and a very small hard disk. They’re designed to work wirelessly over wi-fi or mobile 3G and fit easily into a handbag, shoulder bag or even a pocket. The Dell Mini 9 I’ve been testing might run Microsoft Word okay; I’ve never tried. But it works very well with browsers like Apple Safari and Google Chrome and retails here for just $900; Vodafone will even give you one for free if you sign up to a data plan.

These machines aren’t designed to run desktop apps; they’re all about the Internet. “When Google Chrome comes out over Linux [on netbooks], that’s going to completely change the market,” says Drury. “It will mean that the price of computing is going to drop considerably when an operating system and productivity software isn’t part of purchasing a computer. I don't think it means that Microsoft loses; it means they'll have to change their pricing model as well. That’s when you're going to see the real acceleration of cloud-based services, because everything will be on the web.”

So we have the generation tools, and we have the consumption devices. But what about the content? It seems that content, too, is destined to become ubiquitous, affordable and available. At this year’s Webstock conference in Wellington, founder Adrian Holovaty talked about how his hyperlocal news site was designed and deployed, but said they spend a great deal of their time sweet-talking the public and private bodies that hold the data he wants to share with users: police callouts, real estate listings, health inspections, photos and other local minutiae. There should be a special place in heaven for Holovaty and others who work to unlock and distribute public data: they’re breaking down bureaucratic barriers and creating a true democracy of data. You could say they’re fulfilling the promise of the Internet and the web to bring knowledge tools into the hands of everybody, everywhere, all the time.

Flickr, YouTube and other progressive sites have always made their users’ content available for use on other sites. Unlocking public data is a next vital step. The final step will be when the owners of creative content make it broadly available for use elsewhere, perhaps with a charge involved but importantly without digital locks or crippled features. That’s another political and commercial battle, but once it’s done the barriers to creativity and knowledge will be greatly reduced.

Back in Auckland, Coup and Corkery are planning Koordinates’ push offshore. Their relentless pursuit of map data continues, from rivers in the Chathams to railways in Texas. If they can get the data for free, they’ll share it at no charge; if they have to pay, they’ll clip the ticket and bank some of the proceeds. One thing they don’t have to plan for is adding infrastructure to cope with all that extra data.

“You can do that,” says Coup, “but then the question becomes: is your business about what Jeff Bezos at Amazon described as ‘the muck’, where you’re making sure that disks don’t fail and things are backed up, where data ends up everywhere all the time and whether the servers are even up, and that sort of stuff? Or is the business about actually creating some value and you can let somebody else take care of the muck?”

Magazine layout

Ray Ozzie: cloudless no more

Me 2.0

October, 2008, and Los Angeles seems unusually restrained: the sharemarket is in freefall, the state is near-bankrupt, and a long and arduous presidential campaign is coming to an end. In LA, there’s little obvious politicking going on and little excitement on the street. The Republican Party knows that California is not going to vote for their man and both major candidates are making their pitch elsewhere.

At the massive Los Angeles Convention Centre, however, the crowd is upbeat. Microsoft is holding PDC, Professional Developers’ Conference, an occasional event that sees around 8,000 eager geeks with jobs gather to hear about the Next Big Thing that Microsoft plans to ship.

Ray Ozzie, Microsoft’s chief software architect, unveils the company’s strategy to keep Windows and Office developers ahead as the economy bites. Among a swag of announcements, the biggest, and perhaps the riskiest, is called Azure.

One definition of ‘azure’ is ‘cloudless’, but Microsoft is getting into cloud computing in a big way. It’s been building datacentres at a Google-like pace, and won’t cede the cloud market to Amazon, Google and the smaller cloud providers that have sprung up. They want that business.

Commercially, this can’t have been an easy decision for Microsoft; it became a behemoth on the back of its Windows and Office products—both closely associated with the desktop. At PDC, Ozzie announces new cloud versions of Word, Excel and the other Office apps, but helping to usher in a cloudy future risks the company’s bread-and-butter products.

Still, Azure is a Microsoft product through-and-through. Ozzie announces that Azure will run web applications built with Microsoft’s .Net framework without modification. Sure enough, a demo is run, a checkbox is clicked, and a standard web app is suddenly running in the cloud. It’s a different approach than Google’s cloud offering, Google App Engine, which currently restricts users to its own fast but feature-poor database and two languages, Python and (recently) Java. Most web apps will need to be at least partly rewritten to run on App Engine. But if you already have a .Net web app, it’s ready to run in the cloud right now.

Isn’t Microsoft worried that its cloud customers will end up spending much less than they would on in-house servers and fully-featured desktops running Windows and Office? Scott Wylie, the director of Microsoft New Zealand’s Developer and Platform Strategy Group, says no. “You could say okay, we’re going to take the on-premise model and we’re going to just drive that to its inevitable conclusion—which would be irrelevance, one day. We firmly believe that there’s a place for on-premise software, and it’s going to depend on all sorts of things for the next ten years on decisions where [applications] live. But at the same time we’re building up this infrastructure that is going to give you the choice.”

The first users of Azure, he says, are likely to be pure web businesses. “It’s perfect for people who want to achieve true Internet scale. And as a New Zealand-focused thing, there are a number of software companies who are building software-as-a-service type things. Azure as a platform is aimed at that next generation of entrepreneurs who want to try something. People who are in New Zealand, saying there’s nothing stopping us now … if we’ve got the idea, we can build the thing and market it.

“Before, it was ‘Where are we going to host it? Where are we going to place it so you can access it from the UK and New Zealand at a good response time?’ We’re removed a lot of     those barriers.

“My personal feeling is that next generation of web entrepreneurs won’t accept that because it’s New Zealand it has to be small. If you and I were starting our software company in five years’ time, we wouldn’t consider buying servers. It would be a waste of our capital.”

Matt Cooney visited Los Angeles courtesy of Microsoft