Resume

• 2001

  Bachelor of Science (BS)

  Computer Science

  Bangladesh University of Engineering and Technology

• 3.82

  Verilog

  Assembly

  Doctor of Philosophy (Ph.D.)

  Computer Science

  Texas A&M University

• Linux

  Windows

  HTML

  MySQL

  CSS

  Programming

  Oracle

  PHP

  JavaScript

  Visual Studio

  High Performance Computing

  C++

  SQL

  C

  Matlab

  On the Performance of MapReduce: A Stochastic Approach

  MapReduce is a highly acclaimed programming paradigm for large-scale information processing. However

  there is no accurate model in the literature that can precisely forecast its run-time and resource usage for a given workload. In this paper

  we derive analytic models for shared-memory MapReduce computations

  in which the run-time and disk I/O are expressed as functions of the workload properties

  hardware configuration

  and algorithms used. We then compare these models against trace-driven simulations using our high-performance MapReduce implementation.

  On the Performance of MapReduce: A Stochastic Approach

  Many BigData applications (e.g.

  MapReduce

  web caching

  search in large graphs) process streams of random key-value records that follow highly skewed frequency distributions. In this work

  we first develop stochastic models for the probability to encounter unique keys during exploration of such streams and their growth rate over time. We then apply these models to\nthe analysis of LRU caching

  MapReduce overhead

  and various crawl properties (e.g.

  node-degree bias

  frontier size) in random graphs.

  Modeling Randomized Data Streams in Caching

  Data Processing

  and Crawling Applications

  Exponential growth of the web continues to present challenges to the design and scalability of web crawlers. Our previous work on a high-performance platform called IRLbot [28] led to the development of new algorithms for realtime URL manipulation

  domain ranking

  and budgeting

  which were tested in a 6.3B-page crawl. Since very little is known about the crawl itself

  our goal in this paper is to undertake an extensive measurement study of the collected dataset and document its crawl dynamics. We also propose a framework for modeling the scaling rate of various data structures as crawl size goes to infinity and offer a methodology for comparing crawl coverage to that of commercial search engines.

  Around the Web in Six Weeks: Documenting a Large-Scale Crawl

  High Performance MapReduce

  Shared-memory high performance MapReduce

  much like Phoenix from Standford

  but external-memory and capable of solving MapReduce tasks with arbitrary input size. In this project

  various MapReduce design choices were considered

  developed and bench-marked. The best design (arguable) is the one that uses hash tables for sorting and selection tree for multi-way merge of the sorted runs. While using a large number of threads (on 16 CPU cores)

  the main challenge to scaling is memory-wall. Specially in systems with multiple physical sockets and NUMA latency

  the problem worsens. We work around many such problems to achieve a sorting speed of 332 million keys/s (8 byte keys) which is much better than existing benchmarks.

  Ahmed

  Ahmed

  Structured Data Systems Ltd

  Ranks Telecom Ltd.

  Texas A&M University

  Developed and improved various parts of the telecom billing system. This system consisted of downloading call records from the switching servers

  processing them in an Oracle database using PL/SQL and provisioning customers/numbers. I

  as member of a team

  developed a number of value-added services (e.g.

  top-up account balance

  ringtone and logo downloads etc)

  and the data billing system. I also developed numerous modules of the enterprise management software (PHP/MySQL/Oracle on Linux) used by the employees for their day-to-day job.

  Ranks Telecom Ltd.

  Visiting Assistant Professor

  My research encompasses two fields: characterizing graph crawl experience and high-performance MapReduce systems. \n\nI have developed stochastic models for various aspects of a crawl (e.g.

  degree distribution

  uniqueness of the nodes) on random graphs. By characterizing these aspects

  predictive models for the unseen portion are inferred. In addition

  this has interesting implications in various graph theoretic problems (e.g.

  rumor spreading etc). \n\nMy other research focus is developing and analyzing high-performance MapReduce programs (using C/C++) for multiple-core SMP systems capable of handling arbitrary BigData problems (e.g.

  inverting a 7 TB web graph downloaded by Internet Research Lab

  TAMU). This is complemented by stochastic models for intermediate-data (sorted runs written to disk

  to be merged in the next step) and the total run-time of such programs

  both of which are of significant interest to the BigData community. Over all

  the objective is to examine various MapReduce design/algorithm choices and find out the best mix of them that can process data at the fastest possible rate permitted by the hardware (number of CPU cores and their speed

  memory and I/O bandwidth).

  Texas A&M University

  Software Engineer

  Worked on an existing GIS based software using Visual C/C++.

  Structured Data Systems Ltd

  IEEE

  Member