Making 24/7 eServices a reality: students are always online. Can your Web applications keep up?

Auctions cease at eBay. Amazon.com closes for maintenance. Trading halts at E-Trade. Has hell just frozen over? Relax. There's no need to panic about these fictional events. I'm simply trying to make a point: Whether it's a $15 DVD purchase or a $100,000 stock trade, customers crave reliable Web applications. So do your students: the night owls, early birds, and distance learners who always seem to be online.

In the information age, there's simply no time for colleges and universities to take systems offline for maintenance and backup. Even during acts of God Like blackouts, students demand anywhere, anytime access to your so-called eServices, which include Web sites, digital classrooms, and key systems such as online registration and tuition payment.

START WITH THE BASICS

Yet, how can you build a Web infrastructure that never sleeps? You can start by outlining a High Availability (HA) strategy, whose components include complete network redundancy and "failover" (seamlessly switching from a failed to a live server) mechanisms, "hot swappable" hardware (can be replaced without shutting down a system), and proactive systems management. In a nutshell, you should shoot for at least 99.9 percent availability (or about 9 hours of downtime) per server application per year.

Your data center therefore requires redundant power supplies, multiple Internet and telecom links, and fully tested backup generators. Common sense, right? Oh, and don't forget to have at least two diesel fuel suppliers on call to refill your generators during an extended power emergency. In addition, make sure your broadband, telecom, and power pipelines enter the data center from multiple points around the building. Otherwise, one errant ditch digger can cut all of your voice, data, and power feeds. (And don't think that never happens, because it does.)

Once you've got the basics in place, steal a page from Silicon Valley, and borrow ideas liberally (and legally). IBM created the PC before Compaq Computer cloned it. Xerox crafted the graphical user interface before Apple and Microsoft brought it to the masses. Scores of CIOs have designed IT infrastructures that never fail. Your challenge is finding and emulating the best of them.

Now, I'm not suggesting that you rip off your rival's blueprint for nonstop Web services. Instead, you should mingle with the industry's best CIOs, application service providers (ASPs), technology developers, and consulting teams. You'll be amazed at what Legitimately rubs off on you. But where will you find these wellsprings of top-notch information?

FOLLOW THE MONEY TRAIL

Your first stop should be Wall Street, where key applications simply can't fail. CIOs in the financial services sector are a busy bunch, but they love to talk shop with admirers. If your financial Rolodex is a little thin, however, comb through Web sites Like CIO.com and Informationweek.com to identify Wall Street's most influential CIOs. Once you've found five or six names of interest, pick up the phone and call them. Ask them how their online systems held up during the recent blackout and the Sept. 11 attacks. (Be extra sensitive during the 9-11 discussion, since it remains a delicate subject for many on Wall Street.) Ask them which UPS's (uninterruptible power supplies), backup software, recovery systems, and service providers they trust to keep their IT systems humming along.

During the chats, you might be surprised to hear that only a dozen banks (of more than 500 in the affected area) closed for business the day after this year's blackout, according to The Wall Street Journal. You ask: How is that possible? Answer: Most financial firms have much larger IT budgets and technology staffs than your university.

Still, you can afford to leverage their best practices. One of them is strategic outsourcing. Done right, outsourcing can cut costs and improve reliability. Consider this: Building and maintaining a 99.999 ("five-nines") percent reliable application costs about $380,000 when the project is outsourced, according to Boston-based Yankee Group (www.yankeegroup.com), communication and networking consultants and analysts. In stark contrast, the same project costs about $472,000 when managed in-house.

Skeptical? At times, so am I. We're all far too familiar with outsourcing relationships that flounder or fail. However, there are dozens of reputable outsourcing companies--from Collegis (www.collegis.com) and IBM Global Services (www-1.ibm.com/services) to EDS (www.eds.com)--that specialize in highly reliable applications. If you explore or enter an outsourcing contract, make sure the deal includes clearly defined service level agreements (commonly known as SLAs). Spell out precisely how reliable and responsive you expect your servers and network applications to remain. And use monitoring software to see if your partner is meeting your reliability demands.

SUNSET FOR UNIX?

Many universities, including my own, rely on Unix to run their most critical applications. Unix--particularly IBM's AIX, Hewlett-Packard's HP-UX and Sun's Solaris--provides peace of mind because the core operating system has been refined and enhanced for more than 30 years. Still ... If only Unix wasn't so darn hard on the pocketbook. A highly reliable Unix system typically requires six- or seven-figure hardware.

A more moderate choice is Unix's cousin, Linux. Running on low-cost Intel hardware, Linux now commands about 20 percent of the server market, according to MA-based IT and telecom analysts IDC (www.idc.com).

Linux is quite reliable, but I recommend going the extra mile. Specifically, take some time to explore the Linux-HA (high availability) clustering project (www.linux-ha.orq), which interlinks several Linux servers for improved reliability, availability, and serviceability (RAS). IBM, Intel, SuSE and several other companies have a hand in the Linux-HA project. Their efforts strengthen e-mail servers, databases, firewalls, Web servers and other common applications for your students.

Running Oracle? Be sure to ask the database giant about its "unbreakable" Linux strategy, which is a hit with many customers.

WORKING WITH WINDOWS

By now, you're wondering where Windows fits into your high-availability strategy. I've been a Microsoft server advocate since Windows NT 3.1 debuted in 1993, because nobody recruits developers Like Bill Gates. At some point, I reasoned, NT would gain a critical mass because technology managers wanted a single set of development tools for their servers and desktops.

Fast-forward 10 years and Windows has surged to 55 percent server market share, estimates IDC. Without a doubt, Windows is more reliable than it was a decade ago--yet there's always room for improvement. In fact, Microsoft has issued more than 60 security patches to Windows XP since the operating system's release roughly two years ago. And on the server, Windows remains widely attacked by eager hackers. In October, Microsoft CEO Steve Ballmer vowed to improve the way Microsoft designs and distributes security fixes and software patches for Windows, but it's premature to declare victory.

My advice: Reach out to security software providers like McAfee (us.mcafee.com), Symantec (www.symantec.com) and Computer Associates (www.ca.com) to lock down your Windows systems. Then, comb the market for software that further bolsters Windows' reliability. A company such as Marathon Technologies (www.marathontechnologies.com), which designs fault-tolerance software for Windows, certainly comes to mind.

WHO WANTS YOUR BUSINESS?

There are dozens of additional "reliability" options you could pursue, such as RAID (redundant arrays of inexpensive disks), application load balancing, and grid computing.

But I think your wisest option is to pick up the phone. If fellow CIOs don't return your calls for advice, service providers certainly will. Many are still hurting for money (and customers) in the post dot-com economy. Most will gladly let you tour their data center operations in a quest to get your business. Be sure to take them up on that offer. Then, as you move about a data center, ask about the service provider's strategic partners. Do they work with Cisco Systems, Nortel Networks, or Lucent? Why? Are they a Unix, Linux, or Windows shop? What drove the decision? Listen carefully, take good notes and, if possible, see if the service provider will share its network blueprint with you.

You'll gain Lots of great ideas and, quite possibly, a pre-written high availability strategy that you can leverage in your own university. Clever, huh?

Joseph C. Panettieri is editorial director at the New York Institute of Technology and has covered Silicon Valley since 1992 for publications such as InformationWeek. You can reach him at [e-mail].