I am a computer science student with more that two years work experience.
Currently I’m doing my PhD in Arizona State University, Fulton School of Engineering. My current research is integrating Big-Data systems such as Spark with accelerators like GPU and FPGA. I’m also involved with developing a sophisticated benchmarking tool for OpenCL over both GPU and FPGA, which will demonstrate application requirements for both platforms. Today most data centers are equipped with GPU and FPGA on each node. For example Amazon recently has equipped their nodes with high-end PCIe FPGAs. This provides a great opportunity to utilize these resources seamlessly and provide considerable speed up for various set of workloads such as data analytics. My goal is integrating Spark with OpenCL technology and show how much offloading big-data workloads into accelerators could improve data analysis performance. It also brings several interesting venues to investigate, such as multi-dimensional resource scheduling, data caching on non-host memory, QoS management on multi-resources environment and etc.
My previous experiences include working on Storage Devices in Big Data Systems, Data Center Data Replication Solutions and PCI Express utilization in Data Centers. I have worked and working on several projects, such as enabling Big-Data Systems to support storage and processing of scientific data, Studying the effect of network on the QoS of the applications in Big-Data platforms, and utilizing PCI-Express node to node communication, in order to bring a solid solution for the data replication problem in Data-Centers. I’m also interested in studying the behavior of different storage devices such as SSDs and future 3DxPoint devices on Big Data applications, such as data layering and caching. For a year and 3 months, I’ve been involved as the Founder Engineer in a startup called EITR Systems, which is still in stealth mode and I cannot disclose anything about it yet. In our startup, we were working on data replication in distributed systems and we were able to solve the synchronized data replication into an independent failure domain.
Ph.D. Arizona State University, 2013 - Now
B.S. University of Tehran, 2009 - 2013