CCS 3103: Fault Tolerance

A comprehensive set of vocabulary flashcards covering the principles, terminology, models, and recovery strategies of fault tolerance in distributed systems as presented in the lecture notes.

Fault tolerance

The ability of a distributed system to continue operating properly in the event of the failure of some of its components.

Availability

The property that a system is always ready for use, or the probability that a system is ready or available at a given time.

Reliability

the property that a system can run without failure, for a given time.

Safety

An attribute that indicates the safety issues in the case the system fails.

Maintainability

Refers to the ease of repair to a failed system.

System failure

Occurs in a distributed system when a service cannot be fully provided; it may be partial and can lead to failure escalation.

Fault

A defect within the system.

Error

A deviation from the expected behaviour of the system.

Failure

Occurs when the system can no longer perform as required or does not meet specifications.

Hard (Permanent) Fault

A repeatable error, such as a failed component, power failure, fire, flood, design error, or sabotage.

Soft Fault

A category of faults including transient and intermittent faults as well as operator errors.

Transient Fault

A soft fault that occurs once or seldom, often due to an unstable environment.

Intermittent Fault

A soft fault that occurs randomly where factors influencing the fault are not clearly identified.

Crash Fault Model

A model where nodes simply stop working (fail-stop), which is easy to detect and handle.

Byzantine Fault Model

A model where nodes may behave arbitrarily, sending conflicting or malicious information, requiring sophisticated protocols.

Omission Faults

Failures where messages are lost or omitted, but nodes do not crash.

Timing Faults

Failures where nodes respond too early, too late, or not at all within the expected time.

Fail-Stop Model

A simplified fault assumption where nodes stop and stay stopped.

Amnesia crash

A crash failure where a server halts and loses all history, requiring a reboot.

Pause crash

A crash failure where a server halts but still remembers the state before the crash, allowing for recovery.

Halting crash

A hardware failure where a server halts and must be replaced or re-installed.

Receive omission

A failure where a server fails to receive incoming messages.

Send omission

A failure where a server fails to send messages.

Information Redundancy

The process of adding extra bits, such as parity bits and checksums, to allow for error detection and recovery.

Time Redundancy

A method where an operation is performed and repeated if needed, though it creates delays due to waiting and retransmission.

Physical Redundancy

The duplication of hardware and/or software in the system to improve fault tolerance.

Process resilience

Organizing several identical processes into a group so that if one fails, others can function and service pending requests.

Flat group

A group architecture where all processes are equal and decisions are made collectively, avoiding a single point of failure.

Hierarchical group

A group architecture with an elected coordinator that selects workers to perform operations, allowing for quick decisions.

Two-phase COMMIT ($2$-phase)

A distributed commit protocol involving a VOTE_REQUEST stage followed by GLOBAL_COMMIT or GLOBAL_ABORT based on collective votes.

Heartbeat mechanisms

A fault detection technique where nodes send periodic "alive" signals; missing signals indicate failure.

Rollback Recovery

A recovery strategy that uses checkpoints to restore a system to a known good state.

Forward Recovery

A recovery strategy where the system attempts to move forward by correcting errors using exceptions and exception handlers.

Exception handler

Code declared by a programmer for recovery action to be executed when an exception is raised.