A mathematical model to analyze the utilization of a cloud datacenter middleware

详细信息    查看全文

• 作者：Khaleel Mershad ; ^{kwm03@aub.edu.lb" class="auth_mail" title="E-mail the corresponding author} ; Hassan Artail ^{ha27@aub.edu.lb" class="auth_mail" title="E-mail the corresponding author} ; Mazen Saghir¹ ; ^{mazen.saghir@qatar.tamu.edu" class="auth_mail" title="E-mail the corresponding author} ; Hazem Hajj ^{hh63@aub.edu.lb" class="auth_mail" title="E-mail the corresponding author} ; Mariette Awad ^{ma162@aub.edu.lb" class="auth_mail" title="E-mail the corresponding author}
• 关键词：Cloud computing ; Big data ; Middleware ; FPGA ; Mathematical modeling ; Utilization
• 刊名：Journal of Network and Computer Applications
• 出版年：2016
• 出版时间：January 2016
• 年：2016
• 卷：59
• 期：Complete
• 页码：399-415
• 全文大小：1592 K

文摘

Cloud computing is the future paradigm that will dominate the visualization and processing of web data. Achieving high performance in storing and processing huge amounts of data in the cloud has been a major research concern in the last decade, with the exponential growth of data that has presented itself as a challenge to manage and analyze. The vital element in the success of major cloud businesses and companies depends on their abilities to store Big data and make it available to their users and clients in a satisfactory manner. In a previous work, we presented a system which combines active solid state drives and reconfigurable FPGAs (which we called reconfigurable active SSD nodes or simply RASSD nodes) into a storage-compute node that can be used by a cloud datacenter to achieve accelerated computations while running data intensive applications. In order to hide the complexity of accessing and processing the data stored on these distributed nodes from application developers, we proposed in another work a framework for middleware functionality through an API abstraction layer. The middleware handles all low-level hardware related communications, thus allowing programmers to focus on the application, and not the underlying specialized architecture. The Middleware Server (MWS), which could take the role of a cloud datacenter controller, bridges the connection between a client and the set of RASSD nodes and is a vital component in the viability of the RASSD system. In this paper, we propose a mathematical analysis model to evaluate the performance of the RASSD MWS. We study the utilization of three important elements of the MWS: CPU, memory, and network bandwidth. For each, we derive the parameters that govern and affect its operations, and propose formulas for its utilization factor. We use the analysis results, while applying different values to the system parameters, to illustrate important benefits and limitations of the system.