Transcript Revisiting Distributed Simulation and the Grid

Panel: Revisiting Distributed
Simulation and the Grid
IEEE DS-RT 2005 Montreal Canada Oct. 11 2005
Geoffrey Fox
Computer Science, Informatics, Physics
Pervasive Technology Laboratories
Indiana University Bloomington IN 47401
[email protected]
http://www.infomall.org
1
Major Changes in Grids over last year





Tend to use term Grid not Grid computing due to
importance of data deluge
This change fine for distributed simulation as often
wants to use substantial data (sensor streams)
Web service base better established with all GGF work
based on this and use by all major projects: GT4,
OMII, EGEE, Asia Grids …
Some disputes in detail such as WS-Man v. WSDM but
important principles agreed by major industry players
Key standards are in debate or changed/emerged
during last year
• Essential features of WSDL and SOAP unchanged
2
Benefits from Grids


Grids do not suggest new simulation paradigms but
rather enables access to growing number of grid and
web service enabled resources
Access to the real time data deluge could lead to new
simulation methods using data assimilation more
extensively
• Could enable new simulation areas


Provide supportable software base supplying all
distributed system features not in RTI
Inevitable interoperability of Grid infrastructure could
lead to more competition in simulation infrastructure
development area
3
Barriers to use of Grids





One can build “mediation” interfaces between HLA
and other simulation systems and the Grid
However hard to fully exploit the Grid (e.g. use realtime data in microscopic simulation) as long as RTI and
Grids use such different distributed system
architectures
As Grids and Web services still evolving (and might
even “fail”/”change radically”) not trivial to justify
major investment
“Grid community” from HPCC not from distributed
simulation community
Grid of GiG not “same” as Grid of GGF/EGA/IBM
4
Alternatives to Grids?


For next few years Grids and Web Services are likely to
continue as dominant distributed simulation
technologies
Grids and web services are by my definition synergistic
(Grids build on WS-*) and so use of WS is equivalent to
use of Grids
• Does say there are more or less complex ways of using Grids


Enhancing existing distributed simulation in
“distributed part” likely to have low ROI unless use
Grids
Could wait 5 years to see next “distributed system
revolution” but non WS/Grid systems will get less and
less attractive
5
Why are Grids Important



Grids are important for DoD because they essentially
address DoD’s problem and have made major progress
in the core infrastructure that DoD has identified
rather qualitatively
Grids are important to distributed simulation because
they address all the distributed systems issues except
simulation and in any sophisticated distributed
simulation package, most of the software is not to do
with simulation but rather the issues Grids address
DoD and Distributed Simulation communities are too
small to go it alone – they need to use technology that
industry will support and enhance
6
Some ideas to Remember










Grids replace previous sophisticated distributed object
technologies because industry won’t support DO’s but will
support services
Grids are managed Services exchanging Messages
Grid Services GS-* extend WS-* Web Service Specifications
Web Service container replaces computer
Service replaces process
A stream is an ordered set of messages
Service Internet replaces Internet: messages replace packets
(Sub)Grids replace Libraries
7 out of 9 NCOW Core Enterprise Services CES directly from
Grid Services; metadata, (part of) messaging, collaboration, and
sensors need special attention
RTI should be enhanced with agreed service interfaces and
interoperable RTI-SOA
7