When the time comes to publish your Web pages, you have to make one of the toughest and most important decisions youll ever face: Where should those pages live? Should you set up your own server and handle this yourself? Or should you find a friendly national, regional, or local Internet service provider and let them do it for you?
These questions are good ones, indeed. Answering them will tell you whether you want to buy or build, or perhaps buy now and build later (or even build now and buy later). Stick with us as we take you through the numbers and the reasons why you might choose to roll your own or to get underneath somebody elses Web umbrella.
The most important thing to understand is your overall objectives. If your organization plans to stake out a major presence on the Internet, or views the Web as a key ingredient to its future success and well-being, building your own server might be a natural extension of other plans.
If your organization simply views the Web as yet another way to disseminate news and information along with other media and techniques, your own Web server might add no value. At that point, you could perform some simple analyses and figure out whether building or buying makes more sense.
Point 1: If the Web is a paramount method for communicating with your audience, your own server helps to establish your organization as a legitimate, full-time Web presence.
|For a detailed analysis of the costs and considerations, please refer to Chapter 16 in the book. A quick and dirty formula goes like this: If Web-related costs from a provider average $1,000 a month or more, its worth considering a server of your own. If average costs are less than $1,000 a month, its probably not worthwhile, unless other compelling factors exist. For example, an Internet service provider may not use its Web server as heavily as its clients, yet theyd look kind of lame if they didnt have one!|
When calculating costs, figure at least one-fourth of a system administrators time will be required to run a Web server. Remember also that costs include monthly communications fees, Internet access charges, and maintenance-related costs, plus amortizing hardware and software costs to set the system up. Its not at all unusual for after-purchase costs to add up to ten or more times initial costs over the life of a system.
Point 2: At $1,000 a month or more for Web services, start thinking about building, rather than buying, a Web server.
One key element in determining Web service costs is to include the download fees usually assessed by service providers. Normally, such costs run from two to ten cents per megabyte of data downloaded. This might not sound like much, but stop to crunch some numbers: If your pages are 2MB in size, and you average 30 users per day, that translates into $36.50 to $182.50 for an average month. Raise the amount of data to 5MB, and the costs go from $91.25 to $456.25 for an average month.
None of these costs is prohibitive, though, nor would they add up to enough, including telephone or access line costs, and monthly account fees, to exceed the $1,000 monthly ceiling delineated in this chapters section Counting Your Pennies. Things get expensive, though, when hundreds to thousands of users a day start downloading data. At that point, multiplying costs three- to thirty-fold can be prohibitive. Thats why initial testing and an audience survey are important when making projections.
Point 3: If you expect (or even hope for) lots of traffic, youre better off building your own server instead of buying space on somebody elses. Dont forget to hire an experienced webmaster or train somebody in-house right away, while youre at it!
The principle of parsimony argues that youd better try to limit the amount of data you publish on your Web pages, especially if youre paying by the megabyte for data transferred from your server. Its important to keep an eye on the amount and kind of data that your Web server offers.
If you have numerous large files or images that some, but not all, users find interesting, consider making them available through other means. This could include finding an anonymous ftp site on the Internet and directing users to pick up large files there, or it could mean setting up an e-mail based file delivery system like listserv or majordomo.
Point 4: Keep an eye on the data that users download and try to make large, infrequently accessed files or documents available by other means.
The Web is an ever-changing galaxy of information. Keeping up with change means regular effort and a fair amount of volatility in your Web offerings. Many providers charge extra when you change materials, but even if they dont, the effort of making and testing changes has associated -- and inescapable -- costs.
We recommend that you manage and schedule changes on your server. Plan regular updates and stick to your plan. Gather up incremental and incidental changes in the meantime and apply them when the schedule or external factors say its time. While youre at it, keep tabs on the freshness of your materials, and regularly check your off-server links to other pages and locations for currency and correctness.
If you have loads of content that changes frequently, you might want the extra control over a server that you get with building your own. Convenience and access to your data have value, too, and may influence you to build your own server even if your costs hover below the $1,000 monthly cut-off.
Point 5: Constant, unceasing change is more expensive than planned change. Whether you build or buy your Web server, you have to manage change, or it manages you! Where change is frequent and regular, the convenience of access to your own server can have significant value.
If your organization has a well-planned communications strategy, effective use of the Web can complement other channels of communication -- like the trade press, the news media, and the industry analysts and pundits who follow your industry. In this kind of environment, Web documents reflect and coordinate with documents of many other kinds, including advertisements, collateral, and a range of other corporate publications.
In environments where tight controls over corporate or organizational communications must be maintained, its pretty normal to find Web servers under the purview of a public relations or corporate communications department. Likewise, in environments where internal and external communications are formal and carefully managed, you might even find one set of servers for internal materials (not available to the public) and another set of servers for external materials.
Point 6: In tightly managed organizations, especially those with carefully orchestrated communications, control over the Web server may be an absolute requirement, irrespective of other considerations.
If you decide to build your own server and connect up to the Internet via a 28.8 Kbps modem, you limit the number of users who can access your Web pages at any given time (probably one to three users for each such connection). If, on the other hand, you attach a full 1.44 Mbps T-1 link to your server, you increase the size (and the cost) of your pipe significantly, but youre still limited to under 100 simultaneous users.
Projecting traffic, as it turns out, is not only important to understanding costs, its also important to matching the size of the Web server pipe against the number of potential users. Its wasteful to use a large pipe for low-traffic situations, but it can be catastrophic to provide a small to medium pipe when a tsunami of interest heads your way. Frustrate your users long enough, and they may decide to meet their interests elsewhere.
Monitoring usage and demand is the only way to cope with this situation, but youll be better off if you start with extra, unused capacity, than if your pipes are clogged from the word Go! Because operating your own server gives you the flexibility to negotiate the right-sized Internet connection with a telecommunications company and a service provider, many large-volume operations prefer to do it themselves. Even though the costs can be high, theyre generally lower than if you pay someone else to provide them. If you dont expect to be able to add or expand capacity at a moments notice, you can avoid getting too frustrated when its time to add more bandwidth.
Point 7: Make it easy for the audience to browse your materials, and they do just that. Make it difficult or impossible, and they go away . . . forever!
Companies of many stripes are hungrily eyeing the millions of Internet users as another customer base, ripe for electronic commerce. Today, you can find several ways to conduct business transactions over the Internet, ranging from so-called digital cash to a variety of credit-card handling operations.
We dont want to take on the responsibility of recommending a particular approach (even though several of the principals at First Virtual Holdings are friends of ours). Well just say that several million dollars worth of business was performed on the Web in the first three quarters of 1997, so it's a phenomenon that's not going away any time soon. It is well worth further investigation.
If your company is thinking about adding electronic commerce to its existing sales channels, you want to consider building your own Web server. Issues of control, access to customer information, and managing the details of financial transactions all argue that the best hands on your Web server are your own!
Point 8: If youre thinking about doing business on the Web, you probably want to control your own server, for a variety of good business reasons.
Whether you build or buy Web services, you must clearly understand how your pages operate. This is especially true for forms-handling, or other CGI-related programs that run on the server to handle user information requests or submissions.
Its imperative that the server that runs your pages is compatible with the services that you provide and the programs that you run on your users behalf. This means specifying the kind of httpd implementation you need, and also means understanding the names and versions of surrounding standard services that CGI programs may use (like the clickable image map implementations on NCSA and CERN httpd, for example).
Point 9: When selecting a Web server, compatibility with CGI programs, related libraries, and other collections of widgets and data is a must. Dont build or buy the wrong server!
Finally, you may have to cope with what many people would consider an enviable problem: What happens if your Web pages become the latest rage, and your server gets completely inundated by users trying to avail themselves of your magnificent content?
If that happens, you want to make arrangements for fallback services. In this extreme case, working with a national Internet service provider -- like AT&T, Microsoft Network, MindSpring, AOL, and so on -- is essential. If you can afford what these companies offer, you can buy as much capacity from them as you can pay for.
Point 10: If youre smitten with boundless success, be prepared to suffer (especially in the checkbook) for your fame, but make alternative arrangements with a national provider to avoid the perils of Point 7.
E-mail: HTML For Dummies
Webmaster: Natanya Pitts, LANWrights
Revised -- January 16, 1998