Saturday, September 23, 2006

Small Network Needs

... or...
Every important thing I know I learned from building a small network.

In the latter half of the 1990's I had the luxury of building a network environment from scratch. I cite this as an exceptional circumstance because very rarely does one get to start with a clean blank canvas when implementing a network. Almost always there are existing infrastructures to contend with and legacy technologies to accommodate. I found myself in a circumstance where there was no pre-existing conditions. I was given liberty to grow a network from a single seed. From that seed a small to mid-sized network has evolved but not without learning curves and growing pains. In recounting the development and evolution of this network I hope to emphasize a few very important lessons learned.

  • Small networks require the same infrastructure components that larger networks require.
  • Planning, planning, planning!
  • Networks are a very good example of "Pay-me-now-or pay-me-later"
  • Computers really only do one job at a time well. Only people think they can multi-task.
  • 100% Over-Pull - Anticipate growth!

Small Business Owners beware... Small networks require the same infrastructure components that larger networks require.

This is a statement-of-the-obvious for any Information Systems Manager worth his or her salt. Unfortunately it comes as a HUGE and dissappointing surprise to most small business owners. Too often the refrain is, "All I want to do is get my e-mail and run some kind of simple ERP application." Unfortunately there are three deceptive inclusions in such a statement. First is that there is no such thing as a simple ERP application. Second, getting e-mail is considerably more involved that just dialing into an ISP. Finally, the largest deception by far is the assertion "all I want to do..." These are the famous last words (actually first words) of Feature Creep, the enxorable march of additional functions and "needs" that suddenly appear as soon as the network is established. Feature Creep is almost always rationalized after the fact by saying, "Well, we already paid all this money, we should at least be able to (fill in the blank)."

We will revisit this point ... again and again ...

Planning, planning, planning!

The best way to combat the "all I want to do..." attitude is to determine exactly what you really do want to do. Ask all the questions first - they are cheap at the price and the answers will offer clear direction for the overall project. Start by determining the general responsibilities of your proposed network environment. File Sharing means a central repository, a file Server (1) that network users can get to. Preventing labor from accessing management files means having a user security system. Print Sharing means that both the computers and the printers have to be interconnected. E-mail means that in some fashion the network has to be able to connect the internet. E-mail, in some cases, may require a Server (1). Having a web site means that a Web Server (1) must be accessible by the Internet. Keeping your information safe means having many different layers of security; firewalls, anti-virus programs, anti-spyware programs, user security systems and reporting systems.

Then there are physical issues to address. How will users PCs communicate? Will you install a wired network? This will include distribution devices (switches and hubs). If you implement fiber-optic cabling then suitable converters will be needed. All of these constituent elements will need suitable places to reside. Most often these will be in closets or phonerooms. Network resources need to be physically secure. It does not good to have a strong password when a thief can steal your file server. Even closets and phonerooms can be vulnerable to someone walking in and "jacking" into your network.

Then there is the realistic evaluation of the actual number of users on a network. All too often I have heard something like 'only the main partners and their assistants...' This suggesting that the number of users is relatively small. Serious followup finds that there are 14 partners, each having a personal secretary (14) and an assistant (14). Each office suite shares two printers (14). Seven of the partners use a laptop in addition to their desktop (7). The HR Manager and her secretary are not considered partners but they use a PC(2) and two printers(2). The receptionist, three records clerks and the Building/Security Manager all use PCs (5) and share two printers (2). By my simple count the users PCs total 55 and the printers total 18. So in this example of a Law Office there are 73 connections that have to be made, not counting any of the servers (at least 3 from the above example).

Computers really only do one job at a time well. Only people think they can multi-task.

Almost to a person small business owners have asked, "Why can't a secretary use that server, the one locked in a closet, to type up invoices?" Well, the most obvious answer is, 'You shouldn't lock secretarys in closets.' The real answer is very simple and very dissappointing. Computers are really only single-taskers. The best scenario in a small business setting is one server is assigned one role or responsibility. Many of my counterparts in the industry will take exception with this but my response is just as simple as my rule. If a server "crashes" how many roles or responsibilities would a business operator like to lose for the duration of the crash? If the File Server malfunctions the E-mail is still operational as is the E-Commerce site that runs on the Web Server.

This highlights a paradigm of business that is worthy of note. The network is not the 'thing'. The network is the communication channel that business 'things' use to transfer valuable information. The network is a utility, not unlike water and electricity, that business can make good use of but are not directly dependent on. Networks don't, as a rule, do business - they make business work better. This brings me to the awareness that computing the ROI of a network capital investment is as difficult as determining the value of running water or dependable electricity. On the other hand, an E-Commerce system that increases annual sales by 200% can clearly be evaluated in terms of ROI.

100% Over-Pull - Anticipate growth

When estimating the number of strands of fiber-optic cable to pull between any two given points I apply the '100% Over-Pull' rule. At a minimum I will specify twice as many strands as I anticipate using. More often than not I will specify 300 - 500% over-pull. Experience has shown me that networks grow in the same way that Features Creep. Once a resource such as a network infractructure is established then everybody will find yet another use for it. So where initially 2 of 6 strand are dedicated to the digital network all of a sudden an additional 6 strands are needed for the transmission of CCTV images.

"Wait! You never mentioned anything about CCTV!!! We thought you were talking about computer networks." It is true that I have been focused on computers but again the network is not the 'thing'. In the case of a fiber-optic network infrastructure it can be a resource to any system that can speak "glass" - computers, CCTV, audio, telephone, HVAC controllers, and more.

The same '100% Over-Pull' rule applies to computers and networking. Just like nature abhores a vacuum, business will grow to fill an empty network. Business will grow very quickly to fill an empty network. So it is imperative to anticipate growth. A 300-500% growth rate is not unusual in the first 5 years of a network. This brings us back to my first point - Small networks require the same infrastructure components that larger networks require. (Did I mention we would revisit it again and again... ) Where the real value can be realized is in the correct specification the first time when building a network infrastructure. And of course this correct specification is an out growth of my second point ( Planning, planning, planning! ) married to the spirit of the 100% Over-Pull rule.

Everything I have offered here is about having realistic expectations. Too often under-sizing our needs and objectives leads us to make less than economically valuable business decisions. To be fair, our fathers and their fathers ran businesses with yellow legal pads and Ticondaroga #2 pencils. The supply-and-demand business rules that served them well just don't seem to stand up to JIT thinking and Business@Internet.speed. When we take the first steps out of our father's business paradigm we must be prepared to hit the cyber-space surfing (...hit the ground running...)


William Meloney is an Information Systems Manager for a small mid-western manufacturing concern. The views expressed within are his alone and do not reflect those of anyone else, living or dead. Do not operate heavy machinery. Be sure you can sleep for 8 full hours before taking. If a condition lasting more than four hour occurs contact your physician immediatly.

. . .