A coarse-grained page cache aware multivariate analytical model for the storage performance of a parallel file system

摘要

Huge requirements of emerging Big Data applications combined with the performance impairmentrnof current I/O subsystems pose a great challenge on data storage, management, andrnaccess performance. In order to design efficient storage systems, a clear understanding aboutrnhow factors in the I/O path affect the performance of a data-intensive application is of utmostrnimportance. This paper reports our ongoing research toward addressing this issue, presentingrna coarse-grained page cache aware multivariate analytical model for the performance of writernoperations in a parallel file system. The proposed modelwas developed to reflect the performancernbehavior observed in an extensive experimental effort, in which the impact of 14 parametersrnin the response time and throughput of the OrangeFS was investigated. More than one millionrnexperiments were carried out using four distinct computing infrastructures, providing a detailedrnperformance characterization. Additionally, a thorough evaluation of the proposed model, coveringmorernthan 14 000 scenarios, is reported, discussing both qualitative and quantitative aspects.rnEvaluation results indicate that themodel succeeded in representing the behavior of the parallelrnfile system performance, achieving a Mean Absolute Percentage Error of 39.94%.