Deven Bansod

Logo

Cairns, QLD, Australia (June 2018)

Academic Projects

At Georgia Tech

Map-Reduce Infrastructure [Github]

Implemented a library that provides a simplified implementation for MapReduce framework to process large datasets
  • The library automatically parallelizes and executes the user-defined Mapper and Reducer functions on a cluster
  • Uses gRPC and Protocol Buffers to achieve communication between the framework, Master process and the Mapper/Reducer workers
  • Provides in-built support for fault tolerance and handles slow straggler workers efficiently, maintaining the overall performance of the MapReduce job

Database System Application for SysLog analysis [Github] [Slides]

Designed and implemented a database system to enable the efficient forensic analysis on system call audit logs
  • Designed a database system backed by Neo4j Graph database for efficient storage and forensic querying of Linux audit logs
  • Implemented the online log reduction algorithm proposed by Xu, Zhang, et. al. for the Linux Audit system that guarantees full forward and backward trackability
  • Developed a CLI that supports in-built forensic queries and allows for automatic online ingestion of new logs

Distributed service of store using gRPC and ThreadPool [Github]

Implemented a distributed service of an e-commerce store interfacing multiple clients and vendors
  • Developed a three-layered distributed system comprising of a store with multiple clients and vendors
  • The multi-threaded store handles product queries from the clients, queries the vendors for their respective quotes and serves them back to the client using asynchronous RPC calls
  • Developed a user-space ThreadPool mini-library to handle multiple client requests concurrently

Virtual Machine Memory coordinator & vCPU scheduler [Github]

Developed a vCPU scheduler and Memory coordinator for guest virtual machines using the libvirt API
  • Implemented a user-space vCPU scheduler and a memory coordinator to dynamically manage the resources assigned to different virtual (guest) machines on a host
  • The scheduler and coordinator collect statistics for each guest machine through hypervisor calls and take appropriate actions to balance CPU and memory assignments each time a predefined interval has passed

FindMyAir [Github UI][Github API][Poster][Report]

Demo at: https://findmyair.herokuapp.com

    PE Collective Sensing System (PECSS)

    Working with Dr. Rosa Arriaga, developed a HIPAA-compliant Android application to augment the Prolonged Exposure (PE) therapy for PTSD
    • Aimed at PTSD patients undergoing PE therapy, the Android application passively measures their engagement/stress levels, provides a progress platform, and augments the treatment best-practices
    • The application, designed to be HIPAA compliant, is built on the principle of "data: encrypted at-rest and encrypted in-flight" and also supports an offline mode of operation

    Barrier synchronization for parallel systems [Github][Report]

    Implemented various barrier synchronization algorithms and analyzed their performance on a multicore system, a distributed system and a combination of both
    • Implemented sense-reversing, tree-based, dissemination barriers using OpenMP and MPI parallelization frameworks
    • Developed an OpenMP-MPI combined barrier targetting a distributed system consisting of multicore nodes
    • Evaluated the performance of these barriers against the native OpenMP and Open MPI barrier implementations

    Evaluating Alternative DNS implementations [Report]

    Implemented the domain name service using different transport protocol alternatives: HTTPS, TLS, QUIC, DTLS and DNSCrypt, and analyzed their performance and privacy implications through a set of experiments
    • Implemented or setup the alternative DNS implementations, conducted qualitative and quantitative analysis, and evaluated its viability in the current Internet scenario
    • The experiments considered the metrics like resolution times, protocol overhead, effect of reusing connections, effect of packet loss while also studying the security implications of privacy provided by these protocols

    Approximate Query Processing [Paper]

    Wrote a term paper surveying the area of Approximate Query Processing (AQP) in database systems through its approaches, extensions and applications
    • This survey presents the development of Approximate Query Processing (AQP) over the past couple of decades
    • We cover the online (ex. naive sampling, online aggrgation, distinct couting etc.) and the offline (ex. histograms, sketches, wavelets etc.) aggregation approaches to AQP
    • We review the modern applications of AQP to spatial data and visual analytics and conclude our discussion outlining the future research opportunities and challenges in the field

    At BITS Pilani

    TBD