Notesfoundations🌱 seedling

Reliability

Also: Reliable System · Resilience

A system is reliable if it keeps working correctly (right results, acceptable performance, secure) even when things go wrong. Things will go wrong; reliability is about containing them.

Why it exists

Disks die, code has bugs, and people make mistakes. A system that only works when nothing breaks doesn't work for long at any scale.

How it works

  • What users expect: it does the job, tolerates their mistakes and unexpected use, stays fast enough under real load, and keeps out unauthorized access.
  • The things that go wrong are faults. The system copes through Fault Tolerance: faults happen, but don't become failures.
  • There are three sources of faults: Hardware Faults (random), Software Faults (systematic) and Human Error (the leading cause in outage studies).
  • Security is the exception: breaches can't be undone, so you prevent them rather than tolerate them.

Trade-offs

Gains Costs
user trust, revenue, no data loss redundant hardware, engineering time, operational complexity

Trading reliability away on purpose is legitimate (a prototype, or a low-margin service), but only if it's a conscious choice.

In practice

  • E-commerce: outages cost revenue and reputation.
  • Business tools: bugs cost productivity, and legal risk if they report figures wrong.
  • "Unimportant" apps still carry real responsibility. A corrupted photo library is somebody's family history.

Pitfalls

  • Treating reliability as "nothing ever breaks". That's impossible; the goal is containment and fast recovery.
  • Thinking only critical systems (aviation, power) need it.

Recall

These cards are also in your review deck, scheduled with spaced repetition.