Computer Science Management

Bikash Sharma

CONTACT ** ***** ****.,***# ***-B Cell: 814-***-****

State College E-mail: ******@***.***.***

Pennsylvania-16803, USA WWW: www.cse.psu.edu/ bus145

RESEARCH

INTERESTS Distributed systems performance, power and fault management

Cloud computing

Virtualization

MapReduce

EDUCATION

The Pennsylvania State University August 2007 to Present

PhD candidate in Computer Science and Engineering

CGPA: 3.82/4.0

Advisor: Dr. Chita R. Das

Relevant Courses : Machine learning & Data mining, Distributed Systems, Oper-

ating System, Cloud Computing, Algorithms Design & Analysis, Data Structures,

Programming Languages, Computer Networks, Computer Security, Applied Statis-

tics, Introduction to Regression Analysis, Information Retrieval & Search Engines,

Computer Architecture, Database Systems.

NIT Rourkela, India August 2003 to May 2007

Bachelor of Technology in Computer Science and Engineering

CGPA: 9.01/10.0

Second best computer science undergraduate

PUBLICATIONS

Bikash Sharma, Tim Wood, Chita R. Das, HybridMR: A Hierarchical Scheduler

for MapReduce Clusters, under submission to USENIX HotCloud 2012.

Bikash Sharma, Ramya Prabhakar, Seung-Hwan Lim, Mahmut T. Kandemir,

Chita R. Das, MROrchestrator: A Dynamic Resource Orchestration Framework

for Hadoop MapReduce Clusters, under submission to IEEE Cloud 2012.

Bikash Sharma, Victor Chudnovsky, Joseph L. Hellerstein, Rasekh Rifaat, Chita

R. Das, Modeling and Synthesizing Task Placement Constraints in Google Com-

pute Clusters, in ACM SOCC 2011..

Seung-Hwan Lim, Bikash Sharma, Byung Chul Tak, Chita R. Das, A Dynamic

Energy Management In Multi-tier Data Centers, in IEEE ISPASS 2011.

Seung-Hwan Lim, Bikash Sharma, Gunwoo Nam, Eun Kyoung Kim, Chita R.

Das, MDCSim: A Multi-tier Data Center Simulation Platform, in IEEE Cluster

2009.

Gunwoo Nam, Pushkar Patankar, Seung-Hwan Lim, Bikash Sharma, Chita R.

Das, A Resilient and Collaborative Replacement Framework for Per-Flow Moni-

toring, in IEEE ICDCS 2009.

Bikash Sharma, Applications of Data Mining in the Management of Performance

and Power in Data Centers, Technical Report, Department of Computer Science

and Engineering, The Pennsylvania State University, December 2009.

PROJECTS Research Projects

Resource management in Hadoop MapReduce Clusters

Developed heuristics for e cient allocation of resources in MapReduce clusters

The goal is to improve cluster resource utilization and application performance

Performance and power management of multi-tier data centers

Developed a multi-tier data center simulation platform.

1 of 4

Analyzed the e ects of high speed interconnects like IBA, 10GigE on optimizing

throughput, latency and energy consumption of data centers.

Privacy Preserving Data Mining (B-Tech Thesis )

Implemented various randomization algorithms for data perturbation to preserve

privacy in statistical databases.

Analysis of approximately regenerating the distorted data suitable for secure data

mining applications using cryptography and stochastic techniques.

Web server cache prefetching

Implemented various web server cache prefetching algorithms to determine the

optimal prefetching technique given speci c browser and Internet tra c conditions.

Software Projects

Hospital management system in rural areas

Developed a software system that could enable rural people to query and gather

information on medical status during emergencies like epidemics.

The project was selected as Microsoft best student project in the penultimate stage.

CiteseerX - Scienti c Literature Digital Library and Search Engine

Developed People Search module in MyCiteseerx, personal portal for Citeseerx

members using Java Spring framework and Lucene search engine library.

Instant Messenger

Developed an instant messenger distributed system using Qt cross-platform appli-

cation framework and C++.

Virtual Machine System Security

Analysis of XSM infrastructure in Xen hypervisor.

Implemented policies for the management of security domains in Xen.

Chord

Implemented a distributed lookup protocol, Chord in C

Network Protocols Simulation

Implemented various network protocols like TCP, UDP using C socket program-

ming

User-level thread library

Implemented various user-level thread functionalities in C including synchroniza-

tion and inter-thread communication primitives

Developed a MapReduce-based distributed application using this thread package

INTERNSHIP IBM Research (Bangalore, India)

Research Intern (May 2011 - August 2011)

Fault Diagnosis for Shared Dynamic Clouds

Worked on techniques for e cient and accurate faults diagnosis in virtualized con-

solidated environments like public utility clouds.

Google Inc. (Seattle, USA)

Software Engineering Intern (May 2010 - December 2010)

Characterization of constraints present in Google compute clusters workload

Detailed study of di erent jobs and machines constraints present in Google work-

load. Developed statistical models to characterize these constraints. Implemented

a simulation tool based on Google scheduler to validate these models using real

workload from Google. Based on the characterization models, developed synthetic

workload generator for these constraints.

Indian Institute of Science (Bangalore, India)

Summer Research Intern (May 2006-August 2006)

2 of 4

Awarded Young Engineering Fellowship Program (YEFP) scholarship for summer

internship at IISc, Bangalore (YEFP scholarship is o ered to only twenty students

(computer science undergraduates) annually in India to increase research awareness

in undergraduate students).

Implemented various web server cache prefetching algorithms for web browsers.

Tata Consultancy Services (Mumbai, India)

Summer Software Intern (May 2005-August 2005)

Worked in a team developing the project McKinsey Consultancy, People Link

Developed a client response system module in Java.

RESEARCH

CONTRIBU- Fault manager for clouds Design of CloudPD, a problem detection and diagnosis

TIONS framework for shared virtualized dynamic clouds. CloudPD can di erentiate cloud

related faults (that are manifestations of virtualization techniques like live migra-

tion and dynamic resizing) from application level performance anomalies with high

accuracy and low false positives. CloudPD is a completely automated end-to-end

system for fault detection and classi cation.

Resource management in data analytics systems like Hadoop MapReduce clusters.

The focus is on increasing the resource utilization of the clusters and reduce the

job completion times of MapReduce jobs. The other dimension to it is the de-

sign of a hierarchical scheduler called HybridMR for Hybrid MapReduce clusters

consisting of native and virtualized machines, with an objective to maximize the

QoS of interactive services running alongside with batch workloads like MapReduce.

Modeling and Synthesis of an important cloud workload property called task place-

ment constraints. Analyzed and quanti ed the performance impact of constraints

in terms of a ecting task scheduling delays. Proposed benchmarking algorithms to

incorporate representative task placement constraints and machine properties from

Google clusters in existing performance benchmarks.

Dynamic energy management in multi-tier data centers- Proposed two techniques

to reduce the energy consumption of multi-tier data centers, while ful lling the

user agreed SLA. The rst heuristic uses dynamic provisioning to nd the min-

imum number of servers in each tier to satisfy the SLAs. The second heuristic

works at the granularity of an individual server and determines the CPU speeds of

servers that provides optimal energy e ciency coupled with exploiting sleep state

of servers to achieve further energy savings.

MDCSim: A Multi-tier Datacenter Simulator, a simulation platform for analyzing

the performance and power usage of multi-tier data centers. MDCSim can simu-

late large data centers consisting of thousands of servers and can experiment with

various server cluster interconnects. This simulation framework provides useful

guidelines in designing e cient and cost-e ective data centers.

Improvement in the per- ow monitoring system of networks by preventing the

over ow of ow table. Using a dynamic framework, we estimate the number of

active ows and heavy hitters to reduce the overhead of vast data exchanges across

networks.

AWARDS

Robert M. Owens Memorial Scholarship in Computer Science and Engineering at

the Pennsylvania State University in 2007.

Young Engineering Fellowship from IISc, Bangalore in 2006.

3 of 4

All Orissa ICSE School of Associations Scholarship award.

Microsoft student project award for Hospital Management System in rural areas .

Oracle Certi ed Associate(OCA).

Second prize for best website design competition at the tech.fest of NIT Rourkela.

TECHNICAL

SKILLS Languages: C, C++, Java, Python

Operating System: Linux, Windows, Solaris

Databases: MySQL, Oracle, MS Access

Tools & IDE: Vim, Emacs, Microsoft Visual Studio, Java Netbeans, Eclipse

Others: Socket programming, Shell scripting, R, CSIM, Qt, C++ STL, GDB de-

bugger, Matlab programming, HTML, Latex

TEACHING

EXPERIENCE Teaching Assistant, Pennsylvania State University

Introduction to Digital Systems

Responsibilities included holding student help sessions, grading, presentation of

some lectures on behalf of the instructor.

Teaching Assistant, Pennsylvania State University

Intermediate C++ programming

Responsibilities included managing labs, grading and help sessions.

Teaching Assistant, Pennsylvania State University

Object-Oriented Programming with JAVA

Responsibilities included managing labs, grading and help sessions.

REFERENCES Available on request.

4 of 4

Contact this candidate