TECHNOLOGY: A server crash course

Mar 01, 2000 10:30 PM  By

Are you prepared for unplanned Website crashes?

Well beyond annoyance and frustration, a Website crash can mean serious losses of dollars and cents – and potentially customers as well. In the case of one major catalog site, which requested anonymity, an hour of downtime resulted in a $50,000 hit in lost sales.

In the Internet’s nascent stages, outages were all too common, and there have certainly been the headline horror stories, such as the day in August ’96 when Internet service provider (ISP) America Online crashed, suspending service for 19 hours. Fortunately, network engineers, hardware manufacturers, and software designers have been collaborating on ways to avoid Website outages, or at least minimize the damage.

Regardless of the type of glitch that causes the outages, you must have systems in place to deal with them. According to Rod Murillo, president of Boulder, CO-based Synertech, which designs, installs, and maintains e-commerce sites, there are four types of Website failures: hardware, software, memory, and network. Maintaining your site on multiple servers can help you avoid all four types of Website crashes – or at least provide a suitable back-up, he says.

Fixing the problem

For instance, in the case of a hardware crash, you need to restart the failed server, while a software or memory crash may require you to reinstall one or more applications. Several off-the-shelf and proprietary software applications from host providers can continually monitor and troubleshoot the front- and back-ends of entire e-commerce networks. More often than not, though, glitches such as AOL’s occur due to overloads on servers, regardless of how many users are on the network. That’s why most systems experts encourage Web catalogers – who have considerably less demand than a company such as AOL – to avoid single-server architecture.

One of Synertech’s clients, Sierra Trading Post, a Sparks, NV-based cataloger of outdoor clothing and equipment, gets about 10,000 Website visitors a day. To handle that load – and anticipated Web volume increases for the holidays – last summer Synertech bumped up Sierra’s number of servers from one to four, with three dedicated to the Web application and the fourth to the site’s database. “Those three Web servers work in tandem as a unit to serve all the requests from users who are surfing the catalog,” says Murillo. “If any one goes down, we use a load-balancing router, and requests are then shared by the remaining two servers.” Each server has the capacity to handle such increases, so there’s no disruption of service while the failed server is attended to.

Most high-volume Web catalogers – those processing roughly 1,000 or more orders a day – now have multiple servers, as well as software applications to monitor usage. Yet there are still plenty of single-server catalogers, typically those just entering the e-commerce arena or those companies that aren’t generating tremendous volume. “The first thing we recommend to them is to add a second server,” says David Fry, president/CEO of Fry Multimedia, an Ann Arbor, MI-based e-commerce service provider.

Indeed, “as those types of [single-server Web] catalogers grow and their sales volumes go up, more and more of them will start to think about these things” and what they can do to prevent crashes, says Scott Anderson, vice president of services for Assist Cornerstone Technologies, an e-commerce software supplier in Salt Lake City.

But unfortunately, the time that most think about spending the extra money on additional servers is when that single server goes down. In those cases, you should be able to back up customer data on a tape drive, zip drive, or other storage device so that you can restore the data later. The server should also have a redundancy component that knows to restart itself in the event of a crash.