|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|152778||2018||46 صفحه PDF||سفارش دهید||18861 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Parallel and Distributed Computing, Volume 113, March 2018, Pages 227-249
Big Science and Big Data are concepts emerged from the rather recent trend comprising the use of extensive computational resources in the industry and academy, respectively. In this context, the storage and processing of ever increasing data volumes have catalyzed the popularization of innovative platforms that, differently from traditional DBMSs, present distinct trade-offs between throughput, latency, capacity and consistency to reach higher orders of scalability. However, in this myriad of platforms, how can one judiciously choose the most appropriate for a given application, or choose to endeavor in a new system development? In this paper, we present a continuous metric based on the concept of normalized entropy to classify large-scale storage systems according to their architectural and data-distribution characteristics, as opposed to the traditional approach of static taxonomies. As a result, we expect to contribute to a better understanding of how suitable each system is for any given application requirement, and guide future developments towards the right direction.