Fault Tolerance
Introduced| English | Meaning |
|---|---|
| fault tolerant/fɒlt ˈtɒlərənt/ | fault tolerant |
| redundancy/rɪˈdʌndənsi/ | Additional or repeated data, capacity or components beyond what is strictly necessary, sometimes provided for reliability. |
| single point of failure/ˈsɪŋɡl pɔɪnt ɒv ˈfeɪlɪə/ | A component whose failure can stop the required service because no effective alternative is available. |
Keeping working when parts fail
- A system is fault tolerant 容错的 when it keeps working even if part of it fails.
- The Internet is the famous example — no single failure can bring the whole thing down.
- This is a design goal, not luck.
- The question is how you build it in.
A fault-tolerant system:
Difficulty 1/5 · 100 coins + a small bonus on first completion today
No single failure brings the whole system down.
Having more than one path or copy so a backup takes over is called ______.
Difficulty 1/5 · 100 coins + a small bonus on first completion today
Redundancy is the main way to achieve fault tolerance.
Redundancy is the key
- The main way to achieve fault tolerance is redundancy 冗余.
- That means more than one path, or more than one copy, so a backup takes over.
- Because the network has many routes, packets take an alternate route around a broken link.
- This is exactly why the Internet is fault tolerant.
Packet switching sends packets by independent routes, so the network tolerates a failed link
Fault tolerant, or a single point of failure?
Redundancy (extra paths or copies) makes a system fault tolerant; relying on one part with no backup creates a single point of failure.
A single point of failure is:
Difficulty 1/5 · 100 coins + a small bonus on first completion today
A network with one central router has a single point of failure.
If one link on the Internet breaks, packets cannot reach their destination.
Difficulty 1/5 · 100 coins + a small bonus on first completion today
False. The Internet has many routes, so packets can travel around a broken link; that redundancy makes it fault tolerant.
Single point of failure
- The danger to avoid is a single point of failure 单点故障 — one part whose failure stops everything.
- A network with only one central router has a single point of failure.
- If that router dies, the whole system stops.
- Redundancy removes single points of failure by adding backups.
Why might a small blog accept a less reliable setup than a hospital?
Difficulty 1/5 · 100 coins + a small bonus on first completion today
Designers weigh the cost of redundancy against the reliability needed.
Running a website on two servers in different cities removes the single point of failure of one server.
Difficulty 1/5 · 100 coins + a small bonus on first completion today
If either server fails, the other keeps the site up.
Redundancy has a cost
- Redundancy is not free — extra cables, servers, and backups all cost money.
- Designers weigh the cost of redundancy against the reliability it buys.
One server vs two. A company runs its site on a single server — a single point of failure; if it crashes, the site goes offline. Adding a second server in another city means the site stays up if either fails. It costs twice as much, but for a business losing money every minute offline, the reliability is worth it.
A fault tolerant system keeps working when a part fails, achieved mainly through redundancy — extra paths or copies so a backup takes over. Avoid a single point of failure, one part that stops everything. Redundancy costs money, so designers weigh cost against the reliability it buys.