Redundantization of interdependent program modules for parallel control computing systems: Organization, estimation of fault-tolerance, formalized description

High-performance Software for Desi Gning Complex Cyber-physical Systems on the Parallel Computers

“…The purpose of the article is a method for constructing a high-performance software complex for designing complex systems on parallel computers in full, based on existing mathematical models, methods and algorithms for multilevel stochastic modeling, which are used to calculate the average execution time for specific complex tasks (defined by a set of interrelated jobs -CIJ) parallel computing systems [10][11][12][13][14][15][16][17]. The software package described in the article solves the problem of design in full.…”

“…We will assume that a certain discipline (criterion) K [14] is given for an unambiguous selection of jobs from buffer B for maintenance in the CD, for example, the choice of job with the smallest number, therefore when the CD is released, the jobs according to the given criterion K have the highest priority (dispatch rule) [13][14][15][16][17][18].…”

“…1 at the end of service a 1 (1) belonging to the first layer and pool P1, are ready to perform the job a 2 (2) and a 4 (3) of the second and third layers with P2 and P3 (see Table 1), which enter the system, and in this case directly on the CD. If the first of these jobs was a 4(3) , then the system does not receive new job, since a 5 (5) is the successor of job a 4(3) is not ready for execution (job 3 (3) was not done), and there are only the job of the 2 -nd tier is a 2 (2). If the execution of a 2( 2) is completed (the only one in the P2 pool), then the system can receive the job of the two following tiers (with l = 2), that is, with P3 and P4, but in this case only one job comes a 3(3) -successor of job a 2(2).…”

“…Note that both jobs of the 3rd tier -a 3 (3) and a 4(3) (from P3) have not yet been performed. The System will receive job from the 5th tier (from P5), in particular a 7 (5) and a 8 (5). For the mathematical model in Fig.…”

“…With regard to parallel CS, this problem is called the prediction of the execution time of complex software systems; the latter are usually defined by graph models and are considered as complexes of interrelated jobs (CIJ)tasks and / or their parallel-sequential fragments (subtasks, processes, program modules) [4][5][6][7][8].…”