Hard-workin’ Web sites

Underneath that snazzy home page, is your Web site’s basement filling up with junk? I ask only because I keep hearing reports of companies that have accumulated hundreds of different Web sites, each with thousands, even millions, of pages. Yikes! What a management nightmare. And they’re run by 10 different content management packages, I bet.

The result is a lot of outdated information, dead links and unnecessary storage and support costs. Plus, there are more pages for your Web site visitors to get lost in, as well as stray pages that might be useful to competitors, hackers or even terrorists.

I’m definitely in the less-is-more camp. Here’s my philosophy on Web sites:

  • Speed thrills. Provide fast-loading pages and a fast path from the front door to the shopping cart. People want to get in and get out.

  • Never let your visitors get lost.
  • Never let your visitors see dead links or typos, which sap your credibility.
  • Purge outdated or useless content — ruthlessly. Content has to earn its way onto your Web pages.

I know it’s getting close to autumn already, but maybe your Web sites are overdue for some Spring cleaning. It’s time to shovel out the debris, consolidate, standardise and provide a cleaner Web experience.

One step might be to get control of your content management systems. Another might be to use application performance management tools to help you identify and solve performance problems. In this special report, we also provide tips on how to freshen up a Web site that’s gone stale.

There are some tricks and tools that IT managers use to spot bottlenecks and speed up their sites: Problem: It was a systems manager’s nightmare — an intermittent failure of a business-critical Web application. But instead of fixing the code, the staff just rebooted the server every night. “It was swept under the rug, like so many problems that have temporary fixes that become permanent fixes,” says Eric Jones, senior network engineer at VF Corp, the world’s largest apparel maker.

But Jones should consider himself lucky. A May study by the Business Internet Group revealed that 205 of 315 sites studied suffered application failures that weren’t visible to IT operations. This research underscores the critical problem that stumbling applications — not inadequate bandwidth, pokey processors or even inept users — are crippling performance on Web sites today.

Slow Web software can add up to lost sales for e-commerce vendors and higher costs if performance problems result in missed thresholds in service-level agreements.

Web software was certainly the culprit at VF Corp. The errant server ran an in-house Web application with poorly written Dynamic Link Libraries (DLL) that caused a memory leak. But the DLL failure didn’t happen every time: VF’s business-to-business partners for its Wrangler- and Lee-brand clothing sites would log in and sometimes get an error page, sometimes not. But because the DLL problem didn’t shut down the Microsoft Internet Information Server (IIS), no alert was sent to IT. Fix: Jones says that by using real-time code-analysis software — AppSight from Identify Software — he was able to locate the troublesome DLL, which was ultimately rewritten in its entirety. But that fix took nearly a year to get through in-house decision-making. In the meantime, an IT staffer had to reboot the IIS server every night. “That was the worst-case solution.”

Problem: When a Web site bogs down, users initially blame the network, says Jim Demos, vice president for global network services at Reader’s Digest Association. But they’re usually wrong, he says.

In one instance at Reader’s Digest, a Java application that linked a Web server to a back-end database was causing a Web slowdown.

The application wasn’t releasing the connection between the servers. Because the database server required a set number of connections, once that number was reached, the next link failed to connect and subsequent users sat waiting. Fix: The magazine publisher invested in SuperAgent, a monitoring package from NetQoS, which uncovered the problem with the Java application.

Detecting the source of a slowdown takes some investigative work. Demos notes that site performance is more variable than, say, mainframes and client/server response rates, where subsecond returns are the norm. “But users have had to become a lot more patient in a Web environment,” Demos says. That’s because the complex interdependencies in Web applications provide many opportunities for software to slow down operations.

Problem: Java was at the centre of a similar headache for Jim Struve, assistant manager of information support services at WEA Trust, an insurance and retirement services enterprise for public school employees in the US state Wisconsin. He says a Java applet couldn’t release the connection to a DB2 database on its Web site.

Fix: Struve used a performance-monitoring tool called Fenway. With it, he discovered that the Java software wasn’t the culprit. Rather, the DB2 application was responsible for not releasing the connection to the Java program.

The trouble with fixes

These IT users found products on the market to help ferret out which applications were causing their performance problems. But while there are a lot of products to choose from, they come with their own sets of problems.

For one, these products aren’t meant for novice systems administrators. “I have one caveat,” says Jones, referring to Identify’s AppSite product. “It needs experts to use it, experts who know Windows workings inside and out.”

Another issue is that the monitoring software itself chews up processing cycles on the servers. Mark Rogers, Identify’s vice president for product management, acknowledges that AppSite 4.0 can command 3 to 5 per cent of a system’s capabilities. Jones says this isn’t a problem at VF, but he advises others to be aware before deploying it of the impact a monitoring tool could have on existing hardware.

At VF, AppSite feeds alerts into the Tivoli network management framework, which in turn fires off trouble tickets to the helpdesk. But while that works at VF, it isn’t sufficient as a long-term solution, says Jean-Pierre Garbani, an analyst at Giga Information Group.

“With the Web infrastructure going mainstream, point tools need to better integrate into managed frameworks,” he says.

A view from the top

Some IT managers continue to use just the management consoles of the monitoring tools, because they offer targeted information on areas of concern. Tom Ballard, chief technology officer at Hoovers, a business information provider, uses the “executive dashboard” from ProactiveNet. The software gives him the right amount of high-level information he needs to see if problems exist and then lets him drill down for details. “That way, I can see if the problem is being fixed,” he says.

Harry Nicholos, assistant director for Unix and Web services at a US university, is content to use the console from Resonate. “We haven’t even considered connecting it to OpenView,” which the university uses to oversee its IT operations. Although imperfect, there are tools that will track down the root causes of a Web bottleneck. And by keeping your Web site running at a nice clip, they can save your company money.

Join the newsletter!


Sign up to gain exclusive access to email subscriptions, event invitations, competitions, giveaways, and much more.

Membership is free, and your security and privacy remain protected. View our privacy policy before signing up.

Error: Please check your email address.

More about Giga Information GroupGlobal Network ServicesHooversMicrosoftNetQoSProactivenetSAP AustraliaTivoliVFWEA

Show Comments