Abstract:
With today’s advancing technology, massive amounts of data are being generated by users each day. Processing these data becomes a critical issue because it’s stored in geographically dispersed locations which add difficulties to data transfers. The variety of operations applied on data adds to the complexity of the problem. They range from simple search techniques to image processing and data mining applications that require sorting data, finding relationships among different sets of data, and extracting patterns. General purpose computers are becoming less capable of handling such diversity of applications, that’s why we have introduced the RASSD platform. RASSD stands for “Reconfigurable Active Solid-Sate Drives”; it is our proposed platform for data processing acceleration and transfer using heterogeneous data storage and compute node. A compute node consists of a Xilinx FPGA development board connected to a Solid-State Drive (SSD) storage device. Every group of these compute nodes is connected to a middleware server that serves requests from client computers. In this work, we propose a Linux-based embedded operating system that is configured to run on top of Microblaze soft-processor. The purpose of this operating system, which we refer to as RASSD OS 2, is to provide the needed set of services to the platform. These services include controlling the execution of drivelets (FPGA mini-applications) on the computing node, handling the dynamic reconfiguration of the hardware accelerators, and providing an interface between the Xilinx ML605 FPGA development board and SATA-based SSD. We have developed the RASSD operating system based on Linux to provide the tasks required by the middleware. These tasks include: a TCP server, a TCP client, a TFTP server, a drivelets loader, a dynamic partial reconfiguration manager, and a SATA controller. The TCP server receives commands from the middleware, while the TCP client sends heartbeats to the middleware. The TFTP server communicates data, bit
Description:
Thesis (M.E.)-- American University of Beirut, Department of Electrical and Computer Engineeering, 2013.
Advisor : Dr. Haitham Akkary, Associate Professor, Electrical and Computer Engineering--Committee Members : Dr. Hassan A. Artail, Professor, Electrical and Computer Engineering ; Dr. Mazen Saghir, Associate Professor, Electrical and Computer Engineering, Texas AandM University at Qatar (TAMUQ).
Includes bibliographical references (leaves 84-88)