Perhaps you are challenged with defending your information technology infrastructure budget. The person holding budget authority over the spend plan doesn't understand much about network architectures and computing infrastructure. They are skeptical that the situation is as dire as you make it seem. You explain that the core switches are nine years old, the servers are seven, and the desktops are two. You skipped the switch upgrade two years ago because of a budget crunch and concentrated on upgrading the desktops instead. You intend to replace the switches this year, and the servers next year which should get the infrastructure life-cycle plan back on track.
The budgeteers doesn't agree and thinks you can save money by ignoring your planned switch and server upgrades, and waiting two more years to upgrade the desktops. They note that you’ve rarely had to do any corrective maintenance on the infrastructure, and that ignoring the upgrades are an acceptable risk. To your boss, it looks good on paper. To you, it feels like a disaster waiting to happen. So how do you explain the risks and defend your plan?
Try using Moore’s Law; not in an esoteric sense, but as a way to calculate a kind of cyber “dog years” for your IT infrastructure.
Moore's Law is not an actual law; it’s an observation made by Dr. Gordon E. Moore, co-founder of Intel. It states that the number of transistors on an integrated circuit doubles every two years. David House, also at Intel, noted that the doubling actually occurred every 18 months; hence, the law is quoted that way. Over the years, the observation has largely held true. Many times, market watchers heralded the end of the law as integrated chips approached sizes so small they strained against the laws of physics; but as integrated chip manufacturers reached their limits, new breakthroughs in physics and computer architecture and design helped breathe new life into the law. Moore’s Law has held true now for nearly 40 years.
Using Moore’s Law as a starting point, let’s assume that the number of transistors on an integrated circuit doubles every 18 months as a lower bound and every three years as an upper bound. If we treat the doubling of transistors on an integrated circuit to be the demarcation line for a computing cohort (similar to the way a generation is used in population studies), then a new computing generation (cohort = generation) occurs every one and a half to three years. A human generation is widely accepted as every 20 years (actually 22-37 years), so using a bit of math, we can say that a year of maturity for an integrated chip is equivalent to seven to 13 years of maturity for a human; or more easily, one-human year is equivalent to seven to 13 cyber ”dog years.”
Using 10 cyber-dog years to one-human year as a rule of thumb, we have a more convenient method for communicating the effective age of computing infrastructure. Switches and servers typically have a life cycle of five to seven years in real human time. That means we retire infrastructure after 50-70 cyber-dog years of service, which sounds about right. Infrastructure typically has a fairly constant, noncycling load so can be maintained in service for a long period of time, say similar to a human office worker whose body doesn’t undergo continuous physical stress. In fact, older switches are more prone to fail when you cycle power during routine maintenance.
Desktop computers typically have a life cycle of three years. That means after 10 to 30 cyber-dog years of service, equal to a typical military career, we retire the desktop from active front-line service and it begins a new life working in a public library or school. Interesting side note: using thin clients or virtual desktops makes even more sense since you transfer that desktop replacement cost to the infrastructure. There is good evidence to suggest that in the long run you lower your life-cycle costs as a result, but that’s a discussion for another time.
For the scenario introduced earlier, you can explain to your boss that a nine-year-old switch that has rarely needed repair is like a 90-year-old person in good health; the switch and its component cards may be operating fine now, but they get frailer with age. Any corrective maintenance may prove fatal, especially if that repair requires cycling power. The servers are in better shape, but are also near the end of their life, similar to a working person approaching their seventies. The servers can handle the load fine for now, but will soon be like the switches if you don’t go ahead with the scheduled upgrade. Then follow up with a discussion about expected downtime to repair a server or a switch and the prospect of lost productivity while waiting for the infrastructure to be restored.
A cyber-dog year is equivalent to about seven to 13 human years; use one cyber year to 10 human years as a useful rule of thumb. This may help communicate the costs, benefits, and risks of operating with an aging infrastructure to those in your organization who are not familiar with its design and maintenance. Just hope they don’t counter with anecdotes about the B-52.
U.S. Navy Capt. Ramberto "Rambo" Torruella, N6 for U.S. Naval Forces Europe-Africa / U.S. 6th Fleet (CNE-CNA-C6F).