SCALE

	Contents
	Scale, by Thomas Claburn
	Content Management, by Thomas Claburn
	Security, by Charles Babcock
	Lightweight Development, by Charles Babcock
	The User Experience, by Aaron Ricadela
	Communities, by Aaron Ricadela
	Interactive Timeline: A Brief History Of Web 2.0

Few people launching a web site expect millions or tens of millions of visitors right away. But it happens. Ask YouTube, which had 20.8 million unique U.S. visitors in September, up from 114,000 a year earlier, according to Internet metrics company comScore Networks.

For most sites, network and IT infrastructure doesn't matter as much as you might think. True, there's no business without scalable, available IT resources--but that's where Web 2.0 companies start, not what sets them apart. Despite the sophisticated and massive tech infrastructure behind Google Video, YouTube proved more popular, prompting the deal for Google to acquire its former competitor. Infrastructure is the ante. A winning hand requires something more: innovation, community dynamic, mojo.

In fact, Web 2.0 startups can get out of the gate without their own data centers. Online retailer Amazon is selling slices of its infrastructure to startups that need help getting their computing up to speed. While that infrastructure--servers, operating systems, database software, network connections--is critical, it doesn't add much to the customer experience, says Adam Selipsky, VP of product management and developer relations for Amazon Web Services. And it can be a resource drain: Companies spend about 70% of their resources building and maintaining an IT foundation, Selipsky says.

In contrast to the first wave of Web expansion, when buying million-dollar servers and getting big fast was the mantra, today's 2.0 startups aren't caught up in the nuts and bolts of computing. "We definitely don't see low-level data centers as being core to our business or value proposition," says Don MacAskill, CEO and co-founder of online photo-sharing site SmugMug, which uses Amazon's S3 storage service, a massive array of storage devices linked via storage management software, to augment homegrown infrastructure.

"It's pretty simple, because Amazon does the heavy lifting for us that involves replicating the files among many data centers and storage media," says Chris MacAskill, president and co-founder of SmugMug.

With 18 employees, SmugMug handles 180,000 paying customers and 115 million photos. "We really think our value proposition is the customer experience," Don MacAskill says. "That includes our Web user interface and our customer service."

(click image for larger view) Cute pictures, and they're sitting on an array of commodity storage devices operated by Amazon.com

Scaling customer service may be more difficult than scaling servers, Don MacAskill says. SmugMug has space right next to YouTube in one Silicon Valley data center, and both companies spend time on many of the same infrastructure challenges, like servers and redundant, self-healing file systems, "It's kind of silly that we're all reinventing the wheel rather than someone coming along and commoditizing it," he says.

The basic stuff is a commodity. David Dudas, co-founder and CTO of online video-editing site Eyespot, points to inexpensive, powerful Intel-based servers; bandwidth from multiple providers available at a fraction of the cost of a few years ago; disk storage that's inexpensive and dense (requiring less data center space); and open source software that includes a free, enterprise-class operating system (Fedora Linux), relational database (MySQL), Web server (Apache), and application framework (Ajax).

Eyespot's advantage is its ability to combine pieces into a scalable online video-editing platform. "All the inexpensive hardware in the world will do you no good if you don't know how to put it together correctly, or if your systems break down or grind to a halt at the 50-million-user mark because of poor architecture," Dudas says.

That's a challenge. One of the keys is putting together a system of IT components--such as servers, databases, routers--that can grow independently of one another. Another key is understanding that different media-serving functions--streaming, image serving, Web page serving, databases, and so on--have different resource requirements.

The rent-a-data-center approach to Web 2.0 infrastructure may work for only so long. At some point, "we need to build our own systems," says Arik Czerniak, co-founder and CEO of Metacafe, the third-fastest-growing site on the Net between August and September. Metacafe had 16.6 million unique users and 492 million page views worldwide in September, comScore says. "At this scale, it's a huge technical challenge to make sure your site is up and running," Czerniak says.

Metacafe designed its own software infrastructure--the services, template libraries, metrics, versioning, and monitoring infrastructures, says co-founder and chief product officer Eyal Hertzog. At the same time, the online video site relies on content delivery network Limelight Networks to cache files for more efficient delivery and on Web hosting company RackSpace for server hosting. Metacafe uses the Lamp (Linux, Apache, MySQL, PHP) software stack.

For Metacafe, scaling efficiently means running the site on several hundred servers rather than several thousand. Says Czerniak, "If we were still using the same technology we were using on day one, we would probably need 10,000 servers."

-- Thomas Claburn