Completed Projects


    Incubators

  • The NIMBLE Environment for Statistical Computing

    Graduated Summer 2018

    Fellow: Claudia Wehrhahn

    Many challenges in data science benefit from increasingly sophisticated statistical models. NIMBLE is becoming increasingly popular but is in need of open-source software leadership that drives its adaptation to parallel and distributed infrastructures as well as in-storage computing environments. The incubator team released NIMBLE with its first BNP modeling tools which included more efficient algorithms for the 4 level hierarchical model.


  • Research

  • Mantle: A Programmable Metadata Load Balancer for the Ceph File System

    Completed Fall 2018

    Fellow: Michael Sevilla (PI: Carlos Maltzahn)

    Mantle is a programmable metadata balancer that separates the metadata migration policies from their mechanisms. The features and APIs are implemented on CephFS. The project team used Mantle to study how to manage and migrate file system metadata to improve performance. To achieve better load balancing, the project focused on the overheads of file system metadata protocols. The project lead chose to not continue with the work as an incubator and instead took a position in industry. Although not working on the project full-time, he continues to contribute to the project and support the existing community of contributors.

  • ZLog & CruzDB: Distributed Shared-log for Software-defined Storage

    Completed Fall 2018

    Fellow: Noah Watkins (PI: Carlos Maltzahn)

    CORFU is a fast shared log approach that leverages flash devices. In this project the research team implemented CORFU on Ceph and investigated the benefits of including this log abstraction into software-defined storage, including the management of shared logs across multiple storage tiers. The project fellow graduated in June 2018 and began working in industry, having chosen to not transition this project into an incubator. However, he continued to support contributions to ZLog – including acting as the head mentor for CROSS's 2018 GSoC student Javier Ron. The ZLog team prototyped the dynamic storage in the reporting period with full integration into ZLog in October 2018. Although the project fellow is no longer working full-time on the project, he intends to continue to contribute to ZLog and CruzDB, and support and promote community development around the project.

  • Strong Consistency in Dynamic Wireless Networks to Enable Safe and Efficient Navigation of Autonomous Vehicle

    Completed Summer 2018

    Fellow: Brendan Short (PI: Ricardo Sanfelice)

    Collaboration with the Hybrid Systems Laboratory

    The objectives of this project was to determine consistency requirements of distributed systems with smart storage devices over realistic networks and design algorithms that assure needed consistency, and to develop an open source software to implement the algorithms in the context of safe operation of autonomous vehicles to provide consistent data. It is believed that future services like unmanned-traffic management will depend upon strong consistency for safe operation. These services will often need to be provided over lossy wireless networks with limited bandwidth, where partitions may be frequent. We studied the consistency requirements of distributed systems with smart storage devices over realistic networks. In particular, the distributed systems will implement algorithms that require large amounts of (dynamically changing) data that is available to all systems via a shared log. The problem of navigation of autonomous systems will serve as the prototype application of this research.

  • Memory and Storage Coordinative Lifetime Enhancement with Near-Data Computing

    Completed Summer 2017

    Fellow: Xiao Liu (PI: Jishen Zhao)

    This project aimed to design a DRAM/NVRAM hybrid memory system, which offers scalable performance and resiliency in data center servers and address memory system scalability by developing a DRAM/NVRAM hybrid memory system. The design adopted several gigabytes of DRAM; the main data storage component is NVRAM. As such, scalability issues associated with DRAM -- the performance and energy overhead of refresh and ECC – can be substantially mitigated. The project also explored hybrid memory management mechanisms that provide a unified memory space for persistent and non-persistent data structures.

  • An Efficient C Library for Unum 2.0

    Work ended Summer 2017

    Fellow: Andrew Klofas (PI: Nic Brummell, Carlos Maltzahn)

    The universal number (unum) is a new digital numerical system for computers. Computation with unums enables higher accuracy math by increasing information density. In order to make widespread adoption possible, this project created an open source C library that includes a framework for optimizing its efficiency on different architectures. The project team implemented basic arithmetic operations (addition, subtraction, multiplication, reciprocation, divide) and began adding an unum 2.0 matrix library