SmartGRID Ongoing research work in Univ. Fribourg and Univ. Applied Sciences of Western Switzerland (HES-SO) SwiNG Grid Day, Bern, Nov. 26th, 2009 Ye HUANG PhD Student, [email protected]http://diuf.unifr.ch/people/huangy Pervasive Artificial Intelligence Group, Dept of Informatics, University of Fribourg, Switzerland Grid Group, Dept of Information and Communication Technologies, HES-SO (FR), Switzerland
12
Embed
SmartGRID Ongoing research work in Univ. Fribourg and Univ. Applied Sciences of Western Switzerland (HES-SO) SwiNG Grid Day, Bern, Nov. 26th, 2009 Ye HUANG.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
SmartGRIDOngoing research work in Univ. Fribourg and Univ. Applied Sciences of
http://diuf.unifr.ch/people/huangyPervasive Artificial Intelligence Group, Dept of Informatics, University of Fribourg, Switzerland
Grid Group, Dept of Information and Communication Technologies, HES-SO (FR), Switzerland
2
Outline
Introduction Smart Resource Management Layer Smart Signaling Layer Future Work
3
SmartGRID Overview
Swarm Agent-Based Scheduling framework for Dynamic, Reactive, Interoperable Grid Computing
Developed in: University of Fribourg, Switzerland
http://diuf.unifr.ch/pai University of Applied Sciences of Western Switzerland (HES-SO)
http://gridgroup.hefr.ch
Two Phd students: One is working on the grid scheduling (Ye) The other is working on swarm agent-based resource discovery
(Amos)
4
SmartGRID Motivation
Filling the gap between Grid applications and resources
How to achieve: Using stable grid
applications Large scale of
resources Mixed policies Unstable,
unreliable network
?
5
SmartGRID layered architecture
Loosely coupled layered architecture. Two layers and one internal interface.
Smart Resource Management Layer
Data Warehouse Interface
Smart Signaling Layer
6
Smart Resource Management Layer (SRML)
Nowadays Grids Grid nodes are well controlled by the local facilities Job sharing is generally limited within each grid node
Our Vision Enabling job sharing between distributed nodes is one more step
towards a universal computing infrastructure
Our Goal Serving the grid as a whole, not for individual nodes Without centralized control
7
SRML (2) – MaGate Scheduler
SRML is comprised by a set of fully decentralized MaGate schedulers (original and implemented)
Remote MaGate
LRM
Self-management, matchmaking, behavior logging
Job submission and analysis Agreement based negotiation
for job sharing and other events Job allocation on local resources Adoption mechanism for
external services (SSL, etc.)
Grid Applications
8
SRML (3) - Interoperable scheduling
Reference experiment to prove the idea of Interoperable scheduling Local suited and Local unsuited jobs are submitted to each node Objective: to increase the “Rate of successfully executed Jobs from
the entire grid Community (RJC)”
First results Different policies applied while adopting the SRML and SSL Different results can be obtained with same policies on grids with
different scope A general guideline should give the principle of intelligent