METHOD OF ACCELERATED DATA RECOVERY IN INFORMATION SYSTEM ON MOBILE PLATFORM IN POST-FAILURE MODE

Abstract

The article considers the problem of accelerated data recovery in an information system on a mobile platform (ISMP) in post-failure mode. The relevance of the study stems from the fact that, following an emergency data failure in an ISMP, there is a need to restore the critical part of the data to proper use as quickly as possible, given limited computational, network, and power resources. The aim of the work is to develop a method of accelerated data recovery, which after a data failure provides the selection of the smallest sufficient recovery action based on the control copy and the change log and reduces the duration of the return of the critical data portion to the correct use under current ISMP’s resource constraints. The work formalizes post-failure data recovery through a set of data elements, indicators of a failure, a critical data portion, a set of control copies and a change log. A set of permissible recovery actions is constructed, within which each action is determined by the selection of a control copy, a fragment of the change log, the composition of immediate recovery data and the composition of data for deferred recovery. A rule for selecting the smallest sufficient recovery action and an algorithm for implementing the method have been developed, which takes into account the duration of recovery, the level of residual disruption of the data critical part, the energy resource consumption, and the number of elements for which immediate recovery is provided in the current action. According to the results of the computational study, it was established that the proposed method has an advantage over the basic recovery methods in terms of the recovery duration, the volume of data transmission and reading, the energy resource consumption, the level of residual disruption of the data critical part, and the proportion of successful completion of recovery within the acceptable time interval. The practical significance of the results obtained lies in the suitability of the method for use in onboard and distributed ISMPs, where after an emergency data failure it is necessary to restore the critical part of information base for operation within a short time interval without proceeding to the full recovery of the entire dataset.