Meaning
Verification of data integrity during retrieval involves calculating the number of set bits to confirm consistency with a stored checksum. The use of memory readback parity identifies bit flips caused by ionizing radiation or electromagnetic interference. This check occurs immediately after the data leaves the storage array.
Data Reliability
Redundancy bits are added to each byte to create an even or odd count. High data reliability is maintained by hardware logic that flags a mismatch before the central processing unit uses the value.
Detection Threshold
Reliability of the system depends on the ability to find single bit errors in a memory word. The detection threshold for memory readback parity is limited to an odd number of bit changes. If two bits flip simultaneously in the same word, the parity bit remains valid and the error remains hidden.
Fault Tolerance
Integration of this mechanism into a safety architecture allows the system to enter a fail safe state upon the discovery of corrupted instructions. Fault tolerance is enhanced by combining the parity check with periodic memory scrubbing. This process involves reading every address to ensure that latent errors do not accumulate over time and lead to a system crash.