The formula in words

Every source you name is a branch on its own. When you say that several of them stand behind one dependency, an event attributed to that dependency reaches all of them at once, so those several count as one branch and the rest stay separate. If you named $N$ sources and $m$ of them carry the dependency, an ordinary event still meets $N$ branches and a shared event meets $N-m+1$. The page reads the lower of the two whenever the shared class has any weeks in it at all.

Where the shared share comes from

A worked example

Alpha is closed in weeks 1, 26 and 27. Beta is closed in weeks 1 and 2. Gamma is closed in weeks 1, 40 and 41. All three carry the selected dependency, so their closed week sets meet at week 1 only.

One week converts to $\left\lfloor\frac{10000 \times 1}{52}\right\rfloor$, which is 192 basis points. With $N$ equal to 3 and $m$ equal to 3, the branch count in that week is $3-3+1=1$.

Selecting Same inbound port keeps week 1 but puts two sources behind the dependency, so the whole count becomes 2. Moving Beta's first window to weeks 10 and 11 removes the common week and returns the whole count to 3.

What the whole count is and what it is not

The answer is a whole number of branches because a branch is a whole thing. The event weighted figure in the third table is an average across two classes of week and it is diagnostic. A fractional average is not a fractional source, and neither number measures how often anything works.

What this does not model

It counts the weeks you typed and the membership you declared. It cannot discover a dependency you did not name, including one in a lower tier, a certifier, an owner, a material or a calendar.

It does not determine that any source is qualified or available, and it returns no percentage of anything. A low count is a named question for evidence and it is not a verdict on a source.

Check it yourself

Nine values are checked and the answer changes once, between zero weeks and one. The whole count is the answer and the rational is the working, and a fractional average is not a fractional source. The crossing is still in the first step across bounds of 0 and 52 at a step of 4, because it is the only one there is.

The calculation domain

Enter 2 to 5 sources and 1 to 3 dependencies. Source names allow 24 characters and dependency names 32. Closure windows determine the share, from 0 to 10,000 whole basis points.

Boundary

It counts typed closure weeks and declared membership without discovering unnamed dependencies, determining that a source is qualified or available, or measuring how often anything works.

Source anchors

NIST parallel redundancy model. The first source states that a parallel redundancy model assumes the branches are independent.

United States Nuclear Regulatory Commission on common cause and dependent failures. The second source describes dependent failures defeating redundancy and identifies common cause groups.

Neither supplies a figure, a rate or a table to this page, and nothing on this page comes from anywhere but your own typing.

Ordered calculation trace

1. Build the union of closed week numbers for each source.
2. Keep the sources that carry the selected dependency.
3. Intersect those sources' closed week sets and count the whole weeks in the intersection.
4. Convert that week count with $\left\lfloor\frac{10000 \times w}{52}\right\rfloor$.
5. Count every named source as an ordinary branch and the selected members as one branch in their shared weeks.
6. Return the lowest whole branch count from the classes of week that occur.

Calculation version

1.0.0