Asünkroonsete partiikoodide empiiriline uuring
Date
2019
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Partiikoodid esitleti Y. Ishai, E. Kushilevitz, R. Ostrovsky ja A. Sahai poolt aastal 2004 hajusfailisüsteemide koormusjaotuseks. A.-E. Riet, V. Skachek ja E. K. Thomas demonstreerisid artiklis, et erinevate kestvusega päringud partiikoode kasutavale failisüsteemile põhjustavad ooteaegu. Samas artiklis esitleti ooteaegade probleemi lahendamiseks asünkroonse partiikoodimudel. Käesoleva lõputöö raames loodi kahe partiikoodi mudeli võrdlemiseks süsteemi prototüüp, mille abil simuleeriti mõlemat mudelit kasutavaid failisüsteeme. Simulatsioonide töökäiku kirjeldatakse detailselt. Lõputöös võeti kasutusele asünkroonsel süsteemimudelil uus parameeter, mille abil suurendati süsteemi poolt saavutatavat täidetavat päringumahtu. Teostatud simulatsioonide tulemused visualiseeriti ning tulemusi kirjeldatakse detailselt.
Batch codes were introduced by Y. Ishai, E. Kushilevitz, R. Ostrovsky and A. Sahai in 2004 for load balancing in distributed storage systems. As it is observed in a paper by A.-E. Riet, V. Skachek and E. K. Thomas, varying service times for the user requests could cause long waiting times in the system based on batch code, thus leading to a suboptimal performance. The asynchronous batch code model was introduced as a solution to this problem. In this thesis, to compare the two models, a system prototype was developed which was used to estimate the performance of these models. The constructed system model is described in detail. The thesis introduces a new parameter to the asynchronous system model, which is called "skip distance". The system performance can be improved by optimizing the value of this parameter. The results of the simulations are visualized and explained in detail.
Batch codes were introduced by Y. Ishai, E. Kushilevitz, R. Ostrovsky and A. Sahai in 2004 for load balancing in distributed storage systems. As it is observed in a paper by A.-E. Riet, V. Skachek and E. K. Thomas, varying service times for the user requests could cause long waiting times in the system based on batch code, thus leading to a suboptimal performance. The asynchronous batch code model was introduced as a solution to this problem. In this thesis, to compare the two models, a system prototype was developed which was used to estimate the performance of these models. The constructed system model is described in detail. The thesis introduces a new parameter to the asynchronous system model, which is called "skip distance". The system performance can be improved by optimizing the value of this parameter. The results of the simulations are visualized and explained in detail.