: ..
:
: 112
:
: 2024
: .. // . - 2024. - . 112. - .7-29.
: , , , ,
(.): fork-join queueing system, queueing system, distribution quantiles, copula
: . , . , . ( ) . . ~ . , . , . . , , .
(.): A fork-join queueing system is considered. It is assumed that the service time distribution on all servers has a Pareto distribution. The dependence between the sojourn times of subtasks in subsystems is studied, which is the main reason for the complexity of analyzing such systems. The sojourn time of a task in the system (or the average response time) is the maximum of the dependent random variables of the sojourn time of subtasks in the system. Approximations of the joint distribution of the sojourn times of subtasks are obtained using copula theory. An approach is also proposed for determining the quantiles of the system response time distribution using a diagonal section of copulas. This approach was previously used to analyze a similar system, but with an exponential distribution of service time. However, the main difference between the system under study and the exponential case is that the type of the distribution function of the sojourn time of a subtask in the subsystem is unknown. Therefore, an analytical approximation is used for the quantiles of the response time distribution in the subsystem under the assumption that the distribution of the time of stay of a subtask in the subsystem is approximated by the Frechet distribution obtained earlier. The estimates obtained for the quantiles and copula of the response time distribution show good agreement with the simulation data.
PDF
: 58, : 27, : 18.