Relation of User-Perceived Response Time to Error Measurements on ISDN Data Links

The LAPD link-level protocol was used to interchange messages on a computer-simulated link. The link was subjected to random errors generated with the Gilbert error-producing model. Tabulations were made of the delays encountered in sending and acknowledging messages, and of facility statistics such as error-free seconds. Objectives are 1) provide quantitative relationships between facility measurements and user-perceived performance on a data circuit, and 2) determine which measurements correlate best with observed performance. Major conclusions include: an interval of one second is recommended for facility measurements, as shorter interval measurements such as bit-error ratio are strongly distorted by certain types of errors; error-free seconds (EFS) correlates poorly with link response time performance; a new measure, the sum of squares of errored-second sequences (SQUESS), yields better correlation,