Embed Size (px)
Citation preview
Mixing the Grid and Clouds: High-throughput Science using Nimrod
orIs the Grid Dead?
David Abramson
Monash e-Science and Grid Engineering Lab (MESSAGE Lab)Faculty of Information Technology
Science Director: Monash e-Research Centre
ARC Professorial Fellow
1
Instructions ..• To highlight their science and its success' so
far and how their work utilizes advanced cyberinfrastructure
• Identify potential ITC barriers to ongoing success
• Paint a picture of the future for their research in the next 3 to 5 years and what demands it may create for cyberinfrastrcture
• Identify concrete experiments or demonstrations which will utilize and/or stress the infrastructure within 12-24 months
What have we been doing over the Pacific?
Strengthen Existing and Establish New Collaborations
Work with Science Teams to Advance Grid Technologies and
Improve the Underlying Infrastructure
In the Pacific Rim and Globally
PRAGMA
http://www.pragma-grid.net
A Practical Collaborative Framework
IOIT-VN
PRIME @ Monash
• Engaged in PRIMEsince 2004
• Projects range from bio-engineering, theoretical chemistry to computer science
• Has underpinned long lasting academic collaborations– Publications
– Presentations at conferences
• Undergraduate students without research experience!
MURPA Seminars
The massively increased
bandwidth was transformational.
Quantity begat quality.
Alan Finkel,
Chancellor, Monash Univ
I’ve participated in numerous
video conferences to date but
nothing like this. The quality
was so high that the
experience was almost as if we
were all in the same room.
Students give seminars
A bit about hype …Clouds and Grids and ….
Gartner Hype Cycle 2000
Gartner Hype Cycle 2005
Gartner Hype Cycle 2007
Gartner Hype Cycle 2008
Gartner Hype Cycle 2009
Background and motivation
Introduction
• University research groups have used varying sources of infrastructure to perform computational science– Rarely been provided on a strict commercial basis.
– Access controlled by the users
– High end facilities peer re-viewed grant, usually made in terms of CPU hours
• Cloud computing is a major shift in – provisioning
– delivery of computing infrastructure and services.
• Shift from – distributed, unmanaged resources to
– scalable centralised services managed in professional data centres, with rapid elasticity of resource and service provisioning to users.
– Commercial cloud services
Policy and technical challenges
• Free resources will not disappear!
• Commercial clouds could provide an overflow capability
• Potential – perform base-load computations on “free”
resources,
– pay-as-you-go ser-vices to meet user demand.
• To date, very few tools can support both styles of resource provisioning.
Grid Enabled Elasticity
• Resources maintained by home organisation
• Distinct administrative domains• Unified compute, instruments and data• Middleware layer• Never solved deployment
– See Goscinski, W. and Abramson, D. “An Infrastructure for the Deployment of e-Science Applications”, in “High Performance Computing (HPC) and Grids in Action”, Volume 16 Advances in Parallel Computing, Editor: L. Grandinetti, March 2008, approx. 540 pp., hardcover, ISBN: 978-1-58603-839-7.
• Standards exploded this vision!– Plus a whole load of useless computer
scientists!
17
Cloud Enabled Elasticity
• Home resource expands elastically
• Cloud providers “join” home resource
• Virtual machines deployed on demand
• Scalable infrastructure– Compute– Doesn’t address instruments
and data• Do we still have a whole load
of useless computer scientists?
18
• Grid (Wide Area)– Wide area computing
– Instruments, data
– Security
– File transport
• Cloud (Local Area) – Elastic resources
– Virtual machines (deployment)
• Underpinned by a computational economy!– Abramson, D., Giddy, J. and Kotler, L. “High Performance Parametric Modeling
with Nimrod/G: Killer Application for the Global Grid?”, International Parallel and Distributed Processing Symposium (IPDPS), pp 520- 528, Cancun, Mexico, May 2000
Hybrid solutions
High throughput science with Nimrod
Nimrod supporting “real” science
• A full parameter sweep is the cross product of all the parameters (Nimrod/G)
• An optimization run minimizes some output metric and returns parameter combinations that do this (Nimrod/O)
• Design of experiments limits number of combinations (Nimrod/E)
• Workflows (Nimrod/K)
ResultsResultsNimrod/O Results
22
Antenna DesignDrug Docking
Aerofoil DesignAerofoil Design
Nimrod/K Workflows• Nimrod/K integrates Kepler with
– Massivly parallel execution mechanism
– Special purpose function of Nimrod/G/O/E
– General purpose workflows from Kepler
– Flexible IO model: Streams to files
Authentication
GUI
Vergil
SMS
Kepler
Core
Extensions
Ptolemy
…Kepler GUI Extensions…
Actor&Data
SEARCH
Type
System
Ext
Provenance
Framework
Kepler
Object
Manager
Documentation
Smart
Re-run /
Failure
Recovery
Parameter Sweep Actors
• Using a MATLAB actor provided by
Kepler
• Local spawn
• Multiple thread ran concurrently on
a computer with 8 cores (2 x quads)
• Workflow execution was just under
8 times faster
• Remote Spawn
• 100’s of remote processes
Nimrod/EK Actors
• Actors for generating and analyzing designs
• Leverage concurrent infrastructure
Nimrod/OK Workflows
26
• Nimrod/K supports parallel execution
• General template for search– Built from key
components
• Can mix and match optimization algorithms
A recent experimentResource #jobs completed Total job time
(h:m:s)
μ / σ Job runtime
(mins)
East 818 1245:37:23 91/5.7
EC2 613 683:34:05 67/14.2
A Grid exemplar
Microscopes
ClustersStorage
Visualization
Grid
Middleware
Grid Architecture for Microscopy
ARC Linkage Grant with LeicaRemote control of Leica Microscope from Kepler
Nov 2008
First OptiPortal/Kepler link Feb 2009
First remote control of Leica Microscope in
Germany to Opti-portal in Australia using Kepler
March 2009.
Bird’s eye capture and display
Zooming into area of interest
Image cleanup and rendering
Image cleanup and renderingParallelism for free!
Strawman Project:
Grid Enabled Microscopy Across the
Pacific (GEMAP)?
• Remote microscopes
– Currently Leica
• Mix of Compute Clusters
– University Clusters (Monash)
– NCRIS (APAC grid)
– Rocks Virtual Clusters (UCSD)
– Commercial services
(Amazon)
• Distributed display devices
– OptIPortals
• Cloud time
– Which cloud?
– Who pays?
• Network
– Reservation?
– Who pays?
• Project funding
– Who pays?
• Faculty Members– Jeff Tan
– Maria Indrawan
• Research Fellows– Blair Bethwaite
– Slavisa Garic
– Donny Kurniawan
– Tom Peachy
• Admin– Rob Gray
• Current PhD Students– Shahaan Ayyub
– Philip Chan
– Colin Enticott
– ABM Russell
– Steve Quinette
– Ngoc Dinh (Minh)
• Completed PhD Students– Greg Watson
– Rajkumar Buyya
– Andrew Lewis
– Nam Tran
– Wojtek Goscinski
– Aaron Searle
– Tim Ho
– Donny Kurniawan
• Funding & Support– Axceleon
– Australian Partnership for Advanced Computing (APAC)
– Australian Research Council
– Cray Inc
– CRC for Enterprise Distributed Systems (DSTC)
– GrangeNet (DCITA)
– Hewlett Packard
– IBM
– Microsoft
– Sun Microsystems
– US Department of Energy
37
