4.2 Analyzing failuresFind the weak spot

| Question it asks | How to answer it |
|---|---|
| How many paths from X to Y? | List them by first step, then second. No device twice in one path. |
| Can X still reach Y if this fails? | Cross out the failed link (or every link of a failed device), then look for any path. |
| Minimum failures to cut X from Y? | Try every single failure. If each has a detour, try pairs. Stop at the first size that works. |
| Is there a single point of failure? | Look for one link or device that every path must cross. |
Here's a network where a careless look gets the wrong answer. Every device has at least two connections, so it looks safe. And it is safe against any single broken connection. But look at device M in the middle.
Every path from the left half (H, J, K) to the right half (N, O, Q) passes through M. So if device M fails, it takes J-M, K-M, M-N and M-O with it, and the halves are split. M is a single point of failure, even though no single connection is. That difference between a failed connection and a failed device is exactly what the exam likes to test.
?Wait, what's a "select two" question?
I thought about it. Show me
- Statement I: cross out J-M. H-K-M-N-Q still works. True.
Why?
One path is all you need to show "can still communicate". - Statement II: cross out M and all four of its connections. Nothing joins the left half to the right half. False.
- Statement III: from N, you can go to M then O (N-M-O), or to Q then O (N-Q-O). Any other route would have to come back through M or Q, repeating a device. Exactly two. True.
- Statement IV: cross out N-Q. Q still connects through O-Q, and everyone else is still connected. False.
Why?
A single point of failure has to break something on its own. N-Q doesn't.
Answer: I and III. The lesson of II: a network can survive any single broken connection and still have one device it cannot lose.
Stuck? Get a hint
- Take the statements one at a time and decide true or false for each.
- For II, remember that a failed device takes every one of its connections with it.
- For III, list the ways out of N: toward M or toward Q.
- AConnection J-K
- BConnection M-N
- CDevice M
- DDevice K
Show the answer
C. Device M is on every path from the left half to the right half. Without connection M-N, J can still go J-M-O-Q-N. Without J-K or device K, J still reaches N through J-M-N. Treating a device like a single connection is the trap.
- ANo single failure can stop any device from communicating with any other
- BIf connection C-E fails, E can no longer communicate with A
- CThere is only one path from A to C
- DIf connection A-B fails, A can no longer communicate with C
Show the answer
B. E's only connection is C-E, so losing it cuts E off. A, B, C and D form a loop: A reaches C through B or through D, so the third and fourth choices are false, and the first is false because C-E (and device C) are single points of failure.

When a network is given as a list of connections, draw it before you read the choices. Then, for each statement, physically cross out what failed. Most wrong answers come from imagining a path through something that's already broken.
- Count paths by first step, then second, with no device repeated.
- Minimum failures: test singles, then pairs. The first size that works is the answer.
- A device can be a single point of failure even when no single connection is.
- Select two: judge each statement on its own, then pick the pair.