Be careful what you wish for... you might get it. Tips on how small companies can use the cloud to deal with large spikes in online demand.
"Every small company has dreams of making it big, even if only for 15 minutes," says Mark Sarbiewski, senior director of product marketing at Hewlett-Packard. But what happens when those dreams actually come true?
Sure, you want your business' Web site to get noticed, become popular, and drive business. But what happens when your site experiences a truly massive increase in traffic? Does it simply crash, or does it scale to take advantage of the influx? And how can a small company possibly be equipped to cope with those kinds of spikes?
A lot depends on whether you know the bump is coming or not, and how much warning you have. There's a big difference between gearing up for Black Friday, say, and dealing with a surprise mention on a high-traffic Web site that drives hundreds of thousands of new visitors to your site.
And then there's The Oprah Effect. Get your product mentioned on The Oprah Winfrey show and you had better be prepared for the deluge to your Web site -- you're likely to have a some time to get ready, but not enough to rebuild your infrastructure.
It's a good problem to have, but it's still a problem. You don't want to blow your one big chance to create repeat customers, but according to Jupiter research, 75% of shoppers who can't complete their transactions never come back. That's especially true for smaller, less-well-known brands. "If Amazon struggles for a day, I'll likely give them the benefit of the doubt," Sarbiewski says. "If some unknown site doesn't work, I'm never coming back. The perception is that they're just not up to it."
Of course, it's precisely those small and midsize companies "who don't have deep pool of expertise and resources" to deal with big spikes, says Sarbiewski.
The key is to do some advance planning before the spikes hit:
Do a performance analysis, either on your own or via a service provider. That will tell you what your scalability capacity really is, point out the bottlenecks and reveal the peaks in your current situation. Knowing that information will allow you to plan for spikes, as well as tune up your site to improve its performance as much as possible.
In attempt to standardize this kind of analysis, HP has just launched a new SaaS-based testing product called Elastic Test. The idea is to use Software as a Service techniques to test a Web site's ability to dramatically scale up or down.
From SMBs to global businesses, Web applications need to be able to cope with peaks.
Monitor your systems. Beyond testing, an early warning system can tell you how well your infrastructure is performing, and perhaps even give you a heads up that a spike is coming. System monitoring is also available as a service, Sarbiewski says, via cloud computing.
Monitoring has two components. First, you want to see what your end users are seeing. Sarbiewski says this involves creating a "steady drumbeat of specific transactions to see if the system is slowing down before people start complaining."
Second, you need to watch what real users are doing on your site. It's not enough just to note how many people are coming, you need to note where on your site they go and identify key stress points and early warning signs that you may be reaching your capacity limits.
Monitoring will tell you how much extra capacity you'll need to deal with various scenarios. Do you need 60 virtual machines? Or just 10?
As Forrester's Margo Visitacion put it in a recent Webcast, it's important for Web merchants to monitor and plan to avoid letting seasonal peaks ruin their holiday season. "Don't Scrooge it up," she said.
Monitoring today can help Web sites plan for tomorrow.
Next Page: Dealing With Oprah-Level Traffic Spikes
