Prior CoFlow schedulers approximate the classic online Shortest-Job-First (SJF) scheduling by using a global scheduler to sort CoFlows into multiple priority queues, and a local scheduler at each network port to schedule the local flows of CoFlows in each priority queue using FIFO. Such a division of the scheduling suffers two problems: (1) The flows of a CoFlow may suffer the out-of-sync problem -- they may be scheduled at different times and become drifting apart, negatively affecting the CoFlow completion time (CCT); (2) FIFO scheduling of flows at each port bears no notion of SJF, leading to suboptimal CCT.
Saath is an online CoFlow scheduler that overcomes the above drawbacks by explicitly exploiting the spatial dimension of CoFlows. Our evaluation using an Azure testbed and simulations of production cluster traces show that compared to Aalo, Saath reduces the CCT on average by 1.78x (P90 = 4.50x), which reduces the job completion time on average by 1.46x (P90 = 1.86x).
Guide: Dr. Y. Charlie Hu, Professor, EE, Purdue University.
We are working on designing job scheduling policies in the data center.
Our focus is to improve job completion time.
I worked with a team called NetArch, that works on optimizing and scaling Google’s public network’s infrastructure, my work was focused on prioritizing errors in Internet traffic in the order they should be tackled. During the internship, along with many advanced Google technologies,
Guide: Dr. Raoul Velazco, Professor, TIMA Laboratory
Developed a program for Arduino processor which provides communication via 3G between one of their experimental FPGA board,
kept at remote location, and webserver. Also developed a user friendly website for managing the experiment.
Guide: Dr. T. Venkatesh, Asst. Professor, Dept. of CSE, IIT Guwahati.
Identified features to classify HTTP (and HTTPS) traffic from flows using other protocols. LCPSVM was implemented in MATLAB to do
this classiffication. Upto 8% better classiffication was achieved as compared to normal SVM with these features.
Guide: Dr. Rainer Lienhart, Professor, Institut for Informatik, Universitat at Augsburg.
Automated the process of creating random poselets and positive and negative samples for it, implemented and used procrustes
distance to compare images closeness.
Proposed use of Β(2, 2) distribution for human poselets selection instead of uniform distribution.
Implementation of TCP without Time out. Course Project at Prudue.
Implemented Paging, RMS scheduling, Solaris Scheduling and few other modules on top of Xinu OS . Course Project at Purdue
Guide: Dr. P.K Das, Professor, Dept. of CSE, IIT Guwahati.
An application software, developed in C#, for complete management of a restaurant.