-- ColinLeavettBrown - 2013-09-09

NEP 3 Objectives

  • ATLAS production on the cloud is currently restricted to workloads requiring modest amounts of input and output data, primarily Monte Carlo simulation.
  • Monte Carlo simulation constitutes approximately one third of all ATLAS processing requirements.
  • Our NEP 3 project proposal seeks to implement solutions for the large data requirements of the remaining two thirds of ATLAS workloads; a problem that needs to be addressed by other disciplines.
  • Cloud Scheduler (CS) is a central component in our processing methodology and in our RPI services. We wish to enhance its functionality in several areas:
    • To enable large data applications, CS should become data-aware.
    • Utilization of the OpenStack API to enhance the functionality when working with OpenStack clouds and improve the user experience.
  • In addition, the distributed cloud model we have demonstrated should have image management capabilities independent of any single cloud and that are cloud type agnostic.
  • We seek to implement image management capabilities for the current cloud types that we use but do this in an extensible way so that other cloud types may be supported easily as the need arise.
  • The automation of grid caches to provide deployment on demand and VM access to the "nearest" cache.
  • We want to explore the use of micro VM images to improve image management. It has the potential to reduce image size from gigabytes to megabytes (three orders of magnitude).
  • We see all the above goals as having utility for our processing needs and for general utility.

This topic: HEPrc > ColinLeavettBrown > NEP3Goals
Topic revision: r1 - 2013-09-09 - crlb
This site is powered by the TWiki collaboration platform Powered by PerlCopyright © 2008-2020 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback