News Analysis: Amid the hum and chuff of Web 2.0 this year, a few marquee sites experienced scalability and service problems. What's the problem, and where is the solution?Last week wasn't a good one for the blogosphere in terms of customer service.
If you were unlucky enough to be a TypePad customer, you found that a server crash Thursday night left you locked out of your account. And if you were a del.icio.us user, you had only intermittent access to your shared bookmarks. The site then went down for two days starting Saturday, due to server problems caused by a power failure.
Other well-known "Web 2.0" companies experienced similar problems this year.
Technorati, the blog search engine, was widely derided for offering poor performance and slow indexing.
Access to Bloglines, the most popular Web-based RSS aggregator, was so slow at times that the company admitted that the site's service had "sucked eggs" for a few months.
The problems among these companies led to widespread discussion online, including an argument about the expectation of scalability.
"I think in most cases it's just an example of companies trying to figure out what they have to do to make their services work," said publisher Tim O'Reilly, who helped to create the Web 2.0 meme. "These are small companies. They have to ask, what does it take?"
Scalability and database issues aren't endemic or indigenous to Web 2.0 companies, but the recent spate of problems and the lingering service issues surrounding some popular services lead some critics to ask: Does Web 2.0 scale?
It's Still a Hardware Problem
By most accounts, Web 2.0 companies need the same types of robust and fault-tolerant architecture that Web 1.0 companies had.
"The problems you're seeing here aren't with things you'd call Web 2.0 technologies," said O'Reilly. "The Web 2.0 aspect of del.icio.us is the people and the ideas, not the hardware."
According to O'Reilly, companies need to continue to invest in infrastructure or risk popularizing services that aren't ready for primetime.
But one of Web 2.0's characteristics is the ability to rapidly build and deploy applications with powerful languages. The process wasn't so easy a few years ago, and some developers may yield to the temptation to deploy too early.
"The tools for building applications—mySQL, Apache, etc—have made it so that it's extra easy to put up a compelling application very fast. And then you think it's done," said Adam Hertz, vice president of engineering at Technorati.
On the contrary, Hertz says, companies need to invest even more heavily in infrastructure and technicians.
"One thing that's changed [since Web 1.0] is that we used to scale vertically with big Sun boxes and the like. Now we tend to scale horizontally with smaller, cheaper boxes. That puts a high premium on management."
Technorati has had its fair share of service problems this year.
Hertz attributes those problems to the complex databases and indices that the company manages but admits there's no easy solution to managing complex database-driven services.
From Organic to Commercial
When a high-traffic Web site's infrastructure isn't built with extensive redundancy and fault tolerance in mind, failures can be dramatic.
A power outage in the facility of del.icio.us knocked the service's servers offline last week. The outage corrupted the site's master database.
Founder Joshua Schachter took the site down and posted a note to his blog. Del.icio.us owner Yahoo, which had just bought the company a week prior, assisted with new servers and expertise. The site was back up and running after two days.
Schacter said the service's RAID was not as effective as it could have been.
"This wasn't a scalability problem, it was a hardware and database problem," said Schacter. "But we've been scaling up over time. We've always been pushing the limits of what we can do. A lot of these problems are due to old database technologies that have to work in a certain way."
Schacter said he simply had to reboot the servers, but the database indices took over 12 hours to rebuild.
"Obviously these are some of the things we'll be helping them with," said a Yahoo spokesperson.
Clay Shirky, a prominent technologist and social software proponent, hypothesized that organic services like del.icio.us can sometimes be behind the curve in adapting technologies to support the service.
Organic services, Shirky said, were those that grow based on user contributions and were probably not developed with extensive fault tolerance in mind.
"There's no such thing as a solution to scaling," Shirky said. "What we're seeing is the tension between the organic development of these sites and the surprise of success."
Reacting Gracefully
While critics disagree on the need for scalability, most do agree on the need for corporate transparency.
Not only should a company's applications degrade gracefully, say critics, but the company itself should also react gracefully to known problems with its products.
When Bloglines suffered scalability problems this year, the company posted a notice on its news page: "We're not going to beat around the bush about this. Bloglines performance has sucked eggs lately."
Likewise, del.icio.us tried to keep its users updated with a blog entry. TypePad posted a notice on its Web site.
"People expect the message to come from me," said Schachter, referring to the del.icio.us outage. "Unfortunately, I'm also the guy who's rebuilding the indices. In the future, I'm hoping we'll be able to use the Yahoo infrastructure to manage all this."
Interaction with customers, while not a panacea, can have a profound impact on how they perceive the services, and can help companies manage their customer's expectations.
"The power of these things is fast access and connected pieces of data, and people understand that's a complicated thing," said Gary Stein, a senior analyst at Jupiter Research.
"It's more about how well they react when something goes wrong."
Web Design 
