Aniruddha Bohra

**** ******* **** ****

In nistreams Networks,

Waltham MA, 02451

*** ******* *********,

Phone: 732-***-****

Princeton NJ 08540

abqjnf@r.postjobfree.com

Phone: 732-***-****

http://www.cs.rutgers.edu/ bohra

R ESEARCH Operating systems focussing on the network and storage subsystems, large scale data storage, network

I NTERESTS le systems, multimedia delivery protocols, availability and fault tolerance in computer systems and

networks.

E DUCATION Ph.D., Computer Science, December 2007;

System Architectures Based on Functionality Of oading

Advisor: Prof. Liviu Iftode

Rutgers University, Piscataway, NJ

M.S., Computer Science, Spring 2002;

TCP Server Architecture for SMP-Based Systems

Advisor: Prof. Liviu Iftode

Rutgers University, Piscataway, NJ

B.E., Computer Engineering, Fall 1999

Netaji Subhas Institute of Technology,

Delhi University, New Delhi, India

In nistreams Networks, Inc.

R ESEARCH AND Founding member and Director

E XPERIENCE Network Transport Division

August 2009 Present

Part of a 4 member founding team of a startup, focused on building large scale virtualized CDNs opti-

mized for multimedia delivery over the Internet. The Virtualized CDN leverages existing CDNs and uses

client-side adaptation to deliver sustained high-bandwidth data delivery while reducing cost through load

concentration and distribution. My role in the company is to design and implement the network transport

protocol for a variety of platforms including Microsoft Windows and embedded systems running Linux

and FreeBSD Operating Systems.

NEC Labs America

Research Staff Member,

IP Networks and Distributed Systems

February 2005 Aug 2009

Researcher involved in the HydraStor project, a distributed content-addressable storage system and a le

Jan 2008

Aug 2009 system for secondary storage. The goal of this project is to build a highly scalable, high throughput

storage system built using commodity off the shelf components. The rst generation of this product is

already available in the market.

My role in this project was two fold. First, I worked on improving the sequential throughput for HydraFS

over NFS. HydraFS is a high-throughput le system but some operations, most notably fsync opera-

tions have high latencies. This leads to a very low throughput over NFS. I developed eager-writeback and

page-laundering that tries to avoid issuing NFS COMMIT requests without violating correctness and dura-

bility guarantees. With these modi cations, the throughput increased by more than 4 times for sequential

write workloads, the primary workload for backup systems.

Second, I worked on a QoS and management framework for HydraStor. As part of this framework,

I developed a content centric performance test suite to allow effective performance evaluation. Using

this framework, I designed and implemented a size-based NFS request scheduling framework for the

LinuxOS.

Researcher involved in the Grid Networking (GriN) project, which studies the wide area network pro-

Feb 2005

Dec 2007 tocols to take advantage of spatial and temporal diversity in the Internet. The goal of the project is to

improve transfer times for large point-to-point and point-to-multipoint transfers.

Contributed to the development of a data dissemination framework for high-bandwidth point-to-multipoint

transfers using multiple application-level multicast trees. Techniques to support real-time data dissem-

ination for live media streaming have also been developed. Results reported for experiments using a

wide-area network testbed have demonstrated more than two-fold improvement in performance.

Applications of the technology include Wide-Area File Systems, Wide-Area backups, and online stream

Replication. Two papers related to this project have been published. Additionally, two patents have been

led and are currently under consideration.

Research team member for the resource allocation framework under the next generation wireless mesh

Feb 2005

Dec 2007 project. The goals of the resource allocation is to improve throughput and fairness in Wireless LAN

for spectrum and spatial resources. The key idea is to impose time-slots over the Carrier Sense Me-

dia Access/Collision Avoidance (CSMA/CA) IEEE802.11 MAC framework. These time slots are then

scheduled according to a resource speci c policy.

Contributed to a study evaluating the impact of channel hopping, where clients and access points period-

ically change channels to improve fairness across a dense, unplanned wireless network. Provided initial

system design and implementation for the system and evaluated the performance in the presence and

absence of a malicious client which attempts to jam all communication across the WLAN. By periodic

channel hopping, the system was able to avoid jamming, supporting up to 70% throughput in presence

of a jammer and with minimal loss of throughput without a jammer. A paper discussing the technology

has been published.

Rutgers University

Research Assistant,

Distributed Computing Laboratory

1999 2004

Research team member for FileWall, which of oads the enforcement of le system policies to an exter-

Dec 2005

Present nal le system proxy. This system interposes on the client-server path and implements policies through

message transformation. It also allows administrators to de ne le system policies using a high-level

language without modifying the client and the server. FileWall can be used to de ne monitoring, ac-

cess control, maintenance, and semantic policies which extend the network le systems. Two papers

describing FileWall have been published.

Research team member for Backdoors, a system architecture for non-intrusive remote healing. Back-

Dec 2003

Dec 2005 doors proposes a novel approach for non-intrusive monitoring, recovery, and repair of computer systems

by of oading this functionality to a remote monitor. It takes advantage of Remote Memory Communica-

tion (RMC) which enables an alternative path for a remote monitor to observe and modify the memory

of a remote system. Techniques for remote recovery of internet service sessions and repair of computer

systems have been proposed for Backdoors.

Research team member for Service Continuations, which provides system support for migrating live

Dec 2001

Dec 2003 Internet service sessions. Service Continuations provides support for high-availability in Internet Ser-

vices by migrating TCP/IP connections to alternative servers transparent to the client applications. Web-

services and streaming media services which are critical to clients have been shown to bene t from

the Service Continuation based session migration. A paper describing Service Continuations has been

published.

Worked on of oading TCP/IP processing in a multiprocessor Operating System to improve performance

Aug 1999

Dec 2002 of network servers. Designed an asymmetric Operating System (TCP Server) which dedicated a subset

of processors to network processing. This study identi ed and characterized synchronization overheads

and indirect overheads due to cache pollution in interrupt driven network processing and proposed a

hybrid interrupt-polling based approach. A paper has been published describing the TCP Server design

and implementation.

Bell Laboratories, Lucent Technologies

Summer Intern,

Information Sciences Research Center

May 2000 Sep 2000

Studied the behaviour of various memory allocator (malloc) implementations in the context of a long-

running le system developed at Bell Labs(Hummingbird), and a snapshot based storage service. The

study found that different implementations of malloc result in varying degrees of fragmentation and

the overheads of fragmentation are much worse for long running applications. A paper describing the

research was published.

Rutgers University

Instructor, Computer Architecture

Sep 2003 Dec 2003

Taught the undergraduate course on Computer Architecture for a semester. Was responsible for de ning

the curriculum, teaching lectures, and administering examinations and projects for two sections. The

course involved introducing the fundamental aspects of Computer Systems and its basic building blocks

including hardware and Operating System components.

S ERVICE Reviewer for Architectural Support for Programming Languages and Operating Systems (ASPLOS),

File and Storage Technologies (FAST), Usenix Annual Technical Conference, High Performance Com-

puter Architecture (HPCA), High Performance Distributed Computing (HPDC), ACM Transactions on

Computer Systems, IEEE Distributed Systems Online.

P UBLICATIONS C. Ungureanu et. al. A high-throughput Content Addressable File System 8th Usenix conference on

File and Storage Technologies, FAST 2010 To appear

A. Bohra, S. Smaldone, and L. Iftode FileWall: A Firewall for Network File Systems Proceedings of

IEEE Dependable Autonomic and Secure Computing, DASC 2007, Baltimore MD, September 2007

A. Bohra and L. Iftode Improving Network Stack Concurrency using TCPServers Proceedings of IEEE

Network Computing and Applications, NCA 2007, Cambridge MA, July 2007

A. Bohra, S. Smaldone, and L. Iftode FRAC: Implementing Role-Based Access Control for Network

File Systems Proceedings of IEEE Network Computing and Applications, NCA 2007, Cambridge MA,

July 2007

R. Kokku, A. Bohra, S. Ganguly, and V. Arun A Multipath Background Network Architecture Pro-

ceedings of IEEE Infocom 2007, Anchorage AK, April 2007

V. Navda, A. Bohra, S. Ganguly, and D. Rubenstein Using Channel Hopping to Increase 802.11 Re-

silience to Jamming Attacks Proceedings of IEEE Infocom Minisymposium, Anchorage AK, April 2007

J. Liang, A. Bohra, H. Zhang, S. Ganguly, and R. Izmailov Minimizing Metadata Access Latency in

Wide Area File Systems Proceedings of IEEE High Performance Computing HiPC 06, Bangalore India,

December 2006

F. Sultan, A. Bohra, P. Gallard, I. Neamtiu, S. Smaldone, Y. Pan. Neamtiu, and L. Iftode. Recovering

Internet Service Sessions from Operating System Failures. in IEEE Internet Computing,ICSI-0116-

0804 Special Issue - Recovery-Oriented Approaches to Dependability, March/April 2005

A. Bohra, I. Neamtiu, P. Gallard, F. Sultan, and L. Iftode. Remote Repair of Operating System

State Using Backdoors. Proceedings of First IEEE International Conference on Autonomic Computing

(ICAC 04), New York NY, June 2004

F. Sultan, A. Bohra, I. Neamtiu, and L. Iftode. Nonintrusive Remote Healing Using Backdoors.

Proceedings of First Workshop on Algorithms and Architectures for Self-Managing Systems(Self Man-

age 03), San Diego CA, June 2003

F. Sultan, A. Bohra, and L. Iftode. Service Continuations: An Operating System Mechanism for Dy-

namic Migration of Internet Service Sessions . Proceedings of IEEE Symposium on Reliable Distributed

Systems, SRDS 03, Florence Italy, October 2003

A. Bohra and E. Gabber. Are Mallocs Free of Fragmentation? Proceedings of Usenix Annual Technical

Conference, Freenix Track,, Boston MA, June 2001

T ECHNICAL R EPORTS A. Bohra, S. Rago, and C. Ungureanu NFS Performance of Sequential Write Workloads July 2009,

NEC Laboratories America Technical Report, Submitted for publication

A. Bohra and C. Ungureanu Using System Invariants Technology for HydraStor: A Developer s Per-

spective Dec 2008, NEC Laboratories America, Technical Report

S. Smaldone, A. Bohra, and L. Iftode Implementing Network File System Policies with FileWall Rut-

gers University, Department of Computer Science, Technical Report DCS-TR-605, Nov. 2006.

A. Bohra, A. Baliga, and L. Iftode Orion: Looking for Constellations in Physical Memory. Rutgers

University, Department of Computer Science, Technical Report DCS-TR-569, January 2005.

F. Sultan, A. Bohra, and L. Iftode. Autonomous Transport Protocols for Content-based Networks .

Rutgers University, Department of Computer Science, Technical Report, DCS-TR-479, March 2002.

M. Rangarajan, A. Bohra, K. Banerjee, E. V. Carrera, R. Bianchini, L. Iftode, and W. Zwaenepoel. TCP

Servers: Of oading TCP/IP Processing in Internet Servers. Design, Implementation, and Performance,

Rutgers University Department of Computer Science Technical Report, DCS-TR-481, March 2002.

PATENT A. Bohra, S. Smaldone, and L. Iftode System and Method for Controlling File System Access Filed

A PPLICATIONS Oct 2008

R. Kokku, A. Bohra, S. Ganguly, and R. Izmailov A Multipath Routing Architecture for Background

Transfers Filed March 2006

S. Ganguly, A. Bohra, R. Izmailov, Y. Kikuchi CoDist: Coding Based Distribution of Large Datasets

Filed March 2005

TALKS A Multipath Background Network Architecture. Rutgers Helsinki Workshop on Spontaneous Network-

ing, Piscataway NJ, June 2006

FileWall: Implementing File Access Policies Using Access Context 1st Rutgers / Pierre and Marie Curie

Workshop, Paris, May 2006

Split-OS : An OS Architecture for Clusters of Intelligent Devices, SOSP Work in Progress Session, Banff

Canada, October 2001

Effects of Interrupts on Multiprocessor Servers. New York Metropolitan Area Distributed Systems Work-

shop, IBM Watson Research Center, Hawthorne NY, 2000

Aniruddha Bohra

R EFERENCES Prof. Liviu Iftode

abqjnf@r.postjobfree.com

Rutgers University,

Department of Computer Science

110 Frelinghuysen Road

Piscataway, NJ 08854

Tel: +1-732-***-****

Dr. Cristian Ungureanu

abqjnf@r.postjobfree.com

Department Head,

NEC Laboratories America

4 Independence Way,

Princeton NJ 08540

Tel: +1-609-***-****

Dr. Samrat Ganguly

abqjnf@r.postjobfree.com

Chief Technology Of cer,

In nistreams Networks Inc

116 Village Boulevard

Princeton NJ 08540

Tel: +1-609-***-****

Contact this candidate