Good question. Indeed, some corporations do have the capital, staff, know-how and time frame available to be their own ASP (“Application Service Provider”, or software host). Most companies, though, do not. It’s one thing to put up a server for a few users and quite another to provide company employees with hosted applications on production servers that can’t be down.

At first glance it might not seem like much. A Terminal Server, a high-speed Internet connection, a few user setups, and voila! You are on your way… Don’t worry that your Internet Service Provider (ISP) might lose connection. Power failures never happen. As for server drives, they run almost indefinitely, and besides, a crash won’t cost you much time or data, and backup tapes are SO reliable. Nobody in your firm gets computer viruses. And Mother Nature is never ever a problem to a data center.

Experienced technicians reading this know better, and they have heard it before, over and over. So, just what does it take to host your own applications?

You need to decide how important it is to your organization that your users continue running. If you are a small office and other organizations are not depending on your service, then it is far less important than if you have outside clients running from your center, especially if they are running a mission-critical application (like an accounting or office management system). Even so, a failure means that employees are not working.

Let’s start with the Internet connection. Between 10K and 40K (40,000 bits per second) is an average requirement for software use via the Internet. You can’t use shared lines, such as DSL or cable-modem services, for hosting because you need guaranteed bandwidth (connection volume and speed). If your ISP’s bandwidth is used by enough local subscribers, your users will slow down, and perhaps even be disconnected.

The other thing you cannot do is rely on a single Internet provider. ISP’s periodically lose their connection to the Internet, and thus, so will you. Multiple T-1 lines, which give the best constant access to the Internet, from at least two ISP’s solve the guaranteed connection issue. Routing (programming the data to follow certain paths between systems) becomes critical and an expert is required.

The cost of T-1 lines varies among providers, but expect to pay an average of $800 to $1,000 per month for each. At 30K per user, two T-1’s (3 Mb, or 3 million bits per second) can provide 100 concurrent users with adequate connection.

Next, let’s discuss electricity. Think of the last time the power was off in your home or office for several hours. Now think about dozens, hundreds, or even thousands of clients around the country unable to work because electricty is off in your building. This is an extremely difficult situation, especially since you have no way of knowing how long power will be down during any outage. Most UPS’s (“Uninterruptible Power Supplies”, or battery backups) last only for ten to twenty minutes, if they are used to supply power to just one system. A quarter-hour of backup power will only provide enough time to down your servers properly, saving files from corruption, but not much more than that. So, a diesel or natural gas generator with an automatic cut-over switch is required for any data center that is hosting company applications. The switch detects power failures longer than a few seconds and automatically turns on the generator, which needs to be large enough to service your entire center. Including installation, a natural gas generator and switch could run $15,000 or more.

Redundancy is the name of the game on hardware. Over the years, servers and associated components have gotten cheaper, both in price and, unfortunately, quality. Hard drives, controllers, memory, and even entire servers can go down, and you need fail-safe measures if you need to keep your clients running. You’ll want multiple routers for your multiple T-1 lines, too. If you keep in mind that your servers are no longer there to give access to files, but are running full desktops or applications 24-7-365, the quality and duplication of each component is critical. Name brands, such as Dell, Compaq or Hewlett Packard, sell corporate servers starting at $6,000, but the software and installation can be several times that. Routers, equipment that allows access to and from the Internet and each server, can cost $1,500 to $4,000 each.

Licenses for the operating system and user applications must also be purchased. It is not unusual to invest $600 to $1,000 per user for normal office applications, not including an automation product. Do not pirate software licensing; many illegal installations are caught, and lawsuits and penalties can cost several times the original cost from the vendors. Illegal use of software also gives disgruntled employees a way of attacking your company.

A firewall is as much of a protection as you can install against outside intruders and randomly seeking viruses. A twelve-year-old in Providence has the availability of free utilities which will pick out a system’s identifying address in Los Angeles, test for leaks, and tell the kid how to hack them. If you have ever used personal firewall software, like Black Ice, you know how many hits you get from address and port searching programs every HOUR. Firewall hardware for server rooms, however costs thousands of dollars and requires expertise not easily found.

Should you go with Microsoft Terminal Server, Citrix Metaframe, or stick to HTML applications running on MS IIS with an SQL back end? Do you intend on using a DOS application that will freeze your dual-processor server? Will the unencrypted packets of the RDP protocol mean a lack of security, or is ICA needed to keep nosy ISP server administrators from snooping on your clients’ data?

Obviously, the experience to make these choices and keep the server room running can be the greatest cost. The most overlooked issue in deciding to run your own data center is the necessity for on-staff technicians. This new complex technology is just not stable enough to withstand constant use without assistance and overseeing. Those who think they can call in technicians only when needed will find that they are needed very often, generating even more fees than salaries. If they are not located in-house, your hosting service can be down for hours waiting for repair.

Of course, if you have people working from home or abroad, you’ll need to provide access to technical help nearly 24 hours a day, even on holidays. The convenience or necessity of working away from the office is such a great benefit to companies that once begun, a firm will never want to give up this ability. Unfortunately, printing or data problems can occur whenever someone is online. If a corporate VP is running a projection at 3 am for an eight o’clock meeting and his print queue gets hung up, he will need immediate attention. In accounting, a month-end procedure can take several hours, and sometimes will be started at work and finished at home in the wee hours.

The average network administrator makes in excess of $50,000 per year and call center technicians make over $30K- these numbers can be higher in regions devoid of available expertise. Your staff of computer geeks can easily cost you hundreds of thousands of dollars. But without them, you’ll be out of business.

So, why not do it yourself? I’d rather ask, “Why re-invent the wheel?” If a company has already invested the time, money and other resources necessary to accomplish the goal of hosting applications, then a firm can’t possibly do better than outsourcing to such a company.