Part 5 · 1 chapters · ~10 min

Capacity and Lifecycle

Modelling demand growth from its drivers, finding the binding capacity limit in a shared dependency, whole-life cost where running exceeds building several times over, retention as designed lifecycle, decommissioning planned at the start, and four lifecycle questions for every review.

7

Growth modelling, decommissioning and whole-life cost

the whole life is the design problem
  1. Model growth from the drivers: loans maturing, peaks, trend.
  2. Capacity: find the binding limit, which is often a shared dependency.
  3. Whole-life cost: running a system usually costs several times what building it did.
  4. Retention belongs in the design: hot storage, then cold, then deleted.
  5. Decommissioning is planned at the start.
  6. Four lifecycle questions belong in every design review.
code
// the growth model behind the figure: simple enough to argue with
const monthly = (m: number) => 12_000 * 1.05 ** m;                  // letters per month
const peakPerSecond = (m: number) => (monthly(m) / 30 / 86_400) * 3 * 20;
// month-end: 3× the daily average, compressed into a few busy hours (×20)
peakPerSecond(0);   // ≈ 0.28 letters/s
peakPerSecond(24);  // ≈ 0.90 letters/s: fine for the service, but each letter needs
                    // several signatures on a shared HSM already near its limit at month end
the course, complete
The loan clearance letter has now been through all of it: a need, requirements including the implicit ones, interfaces with owners, an FMEA and a fault tree, a trade study, and a lifecycle plan. None of this took code, and all of it changes the code that gets written. That is the systems engineering argument in one example.
CAPACITY, LIFECYCLE AND WHOLE-LIFE COST
modelling growth, planning the end, and counting what a system costs after launch
swipe the figure sideways, or tap expand for full screen
1/6
growth model
Growth modelling: letters issued per month follow loans maturing: 12,000 a month today, growing with the loan book at about 5% a month, with peaks at month end and after salary days. A simple model projects about 21,500 a month in a year and 38,700 in two, with month-end peaks at three times the daily average.