Shaw's Milestones and Research Questions

From Bobs Projects
Jump to: navigation, search

Contents

TPR

ShawTPR

Annual Plan

There are incoming needs for energy efficiency, cost saving and heat dissipation, triggered by prevalence of data centres, cloud based services and green notion. Fast Array of Wimpy Nodes - Key Value(FAWN-KV) system, which is coupled with low power processors and Solid State Drives (SSD), exhibits a promising perspective of addressing the issue. However other than naturally parallel tasks, this structure appears to be a bit disappointed in both performance and efficiency. The research question arises that such structure is a viable solution to efficient green computing or is a fancy toy brick? The project also intends to explore - what applications would be suitable for the platform? / Can we extend the adaptation of the platform to act as a more capable unity? For the next 12 months, I propose to conduct the following activities:

  • August, 2011
    • Primary Milestone: Implement FAWN Key-Value store
    • Secondary Milestone: Evaluate the performance of the Atom-based system
  • September, 2011:
    • Primary Milestone: Implement JouleSort on the Atom platform
    • Secondary Milestones:
      • Analyse power consumption and time consuming part
      • Implement ACPI to further save power consumption
      • Build the power efficient JouleSort system base on the Atom platform
      • Submit the result to JouleSort competition aiming for next April
  • October, 2011: Connect the Atom platform through Power over Ethernet
    • Primary Milestone:
    • Secondary Milestones:
      • Explore the techniques that facilitate the management of computer system
  • November,, 2011:
    • Primary Milestone: Implement GA on the Atom platform
    • Secondary Milestones:
      • Apply scientific computation to the Atom platform
      • Evaluate performance and power consumption patterns
  • December, 2011:
    • Primary Milestone: Implement Hadoop on the Atom platform
    • Secondary Milestones:
      • Evaluate the file system
      • Evaluate MapReduce tasks performance
      • HDR monitoring
  • January, 2011:
    • Primary Milestone: Apply Machine Learning algorithms to the Atom platform
    • Secondary Milestones:
      • Symbolic machine learning algorithm
      • Stochastic machine learning algorithm
  • February, 2012:
    • Primary Milestone: Continue Machine Learning implementation
  • March, 2012:
    • Primary Milestone: Implement stream processing tasks
    • Secondary Milestones:
      • Video compressing
      • Log file processing
  • April, 2012:
    • Primary Milestone: Synthesise and compare the results of applications
    • Secondary Milestones:
      • Evaluate the performance of the platform in terms of power consumption, overall performance and scalability
      • Find out the drawbacks and limitations
  • May, 2012:
    • Primary Milestone: Review my output and performance in the previous year and refine research goals and objectives.
    • Secondary Milestone:
      • HDR monitoring
  • June, 2012:
    • Primary Milestone: Preparing the paper regarding to performance and power analysis
    • Secondary Milestones:
      • Answer the question that building the system out of low power devices will still be a efficient solutions?
      • What applications are suitable for the platform?
  • July, 2012:
    • Primary Milestone: Survey regarding similar platforms
    • Secondary Milestones:
      • Similarity and comparison
      • Submit the paper aiming for conference interested in parallel computing.

Research Questions

  • Will the system built out of low power devices still be a power efficient option compared to conventional commodity cluster?
  • What techniques could facilitate managing numbers of nodes within the system?
  • What metrics can be used to effectively quantify the usefulness, merit and value of a low power cluster?
  • What applications this novel platform will be capable of in a scalable manner?
  • What suggestions could be given to construct an application oriented power efficient system?

Comments from the Wise

Any and every criticism/clarification, I/O and feedback is very much appreciated.

Alistair

Eric

Bob