Cloud computing: Supercomputers for hire
Cloud computing can provide all the processing power your company needs for little money, but be ready for rough weather.
(Fortune Small Business) -- For years, the computers at TC3 Health were perfectly happy. The company, based in Costa Mesa, Calif., helps health insurance providers avoid overpayments by flagging billing errors, duplicate payments and identity fraud. Its custom-built software, dubbed TC3 Funnel, running on the company's four servers, could swiftly sift through thousands of health-care claims every day.
But that was before TC3 Health's sudden growing pains. The company started allowing clients to submit their electronic archives - years and years of claims - for retroactive processing. The concept caught on fast. Within a week the company was receiving hundreds of millions of claims every day.
The problem was that TC3 Funnel could handle no more than 1 million claims a day.
Accommodating the increased demand would have cost $750,000 in extra servers and software licenses - plus more than $30,000 a month in new maintenance and hosting fees. "The economics were out of whack," says Paul Horvath, TC3 Health's chief technology officer. "We simply weren't in a position to spend that kind of money."
Horvath found his solution in the clouds - or rather, in cloud computing.
One of the fastest-growing Web industries, cloud computing lets you rent additional storage space and processing power over the Internet without your IT guy having to wheel more million-dollar machines into the server room. You pay as you go, hooking up as many servers as you need for as long as you need them.
In geekspeak, you "reach into the cloud" - provided you're prepared to take the risk of leaving your data there.
Once the exclusive domain of government agencies and research labs, cloud computing is starting to emerge as a mainstream alternative to in-house storage and data processing. For the past three years, Amazon (AMZN, Fortune 500) has offered customers two cloud computing services: Simple Storage Service (S3), for unlimited hard-disk space, and Elastic Compute Cloud (EC2), for on-demand processing. This past October, Microsoft (MSFT, Fortune 500) unveiled Windows Azure, its new cloud computing operating system. Google (GOOG, Fortune 500) has started letting developers build Web applications using its massive server farms. And a host of smaller cloud computing players are springing up, including GoGrid, Nirvanix and The Planet (which boasts that it is growing at twice the rate of Amazon).
Horvath opted for Amazon's EC2 primarily because he needed more processing power, and EC2 was the largest and most established service. Now TC3 Health spends a mere $600 a month on EC2 fees, including customer support. That's $29,400 a month less than what it would have spent on physical servers. The company's revenue rose to $20 million last year, which might never have happened had TC3 not reached into the cloud.
Cloud computing services secure data through user authentication, encryption and firewalls. But there's a catch.
"There are legitimate questions enterprises should ask about the security, scalability, availability and reliability of a cloud computing solution," says John Sloan, an analyst with Info-Tech Research Group in London, Ontario. Hackers have yet to siphon data out of a cloud, but the services themselves have experienced some serious outages.
For all the cost savings, TC3 Health did run into deployment challenges. Because TC3 Funnel runs on Windows, and EC2 didn't support Windows servers - not until last October, at least - the company had to retrofit its claims-management system using a specialized application called RightScale. And for security reasons, Horvath modified TC3 Funnel to purge confidential health information in the claims before delivering them to Amazon's cloud for processing. All told, TC3 Health invested six months' time and $150,000 in modifications and precautions.
Satoshi Nakajima was less prepared for the thunderbolt that cloud computing hurled at him. Formerly the lead software architect on Microsoft Windows, Nakajima founded Big Canvas, a startup based in Bellevue, Wash. that develops applications for Apple's (AAPL, Fortune 500) iPhone. The company's flagship product, PhotoShare, allows users to swap iPhone photos for free.
Rather than invest $25,000 in servers, Nakajima paid $900 a month for Amazon's EC2 and S3 services. Big Canvas parked its 50,000 users' photos on S3. Then, last July 20, Amazon S3 went down for seven hours - the service's second outage in 2008.
"It was entirely unexpected," says Nakajima. "And we were upset."
Amazon's explanation: Its servers had stopped talking to one another correctly and had to be taken offline to fix the trouble. By way of a refund, Amazon didn't charge Big Canvas for those seven hours of downtime - a discount of less than 1% on its monthly bill.
To avert a complete disaster, Nakajima quickly tweaked the company's EC2 server to temporarily hold users' photos before transferring them back to the S3 server. But some damage had already been done. By the time S3 came back online, Big Canvas had lost 50 customers' photos - and its pride.
"We had to call all 50 users to apologize," Nakajima says. Big Canvas's EC2 server is now configured to act automatically as a temporary storage server in the event of another S3 outage.
For some users, such glitches highlight the importance of the provider's customer service skills. That was the case for FreshBooks, a Toronto-based online invoicing and time-tracking service. The 30-person outfit recently signed on to Cloud Files, which is available through a subsidiary of major cloud computing provider Rackspace.
"I can get someone on the phone at Rackspace," says Mike McDerment, FreshBooks' CEO. "I can't even find a phone number on Amazon's Web site."
Of course, the average customer service rep can't do much to ensure system compatibility or stave off the outages to which cloud computing is vulnerable. But if you back up your data for the occasional rainy day, these affordable services could send your bottom line sky-high.