User Tools

Site Tools


pads:artis

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision
Previous revision
pads:artis [2011/06/30 13:42] gdangelopads:artis [2014/01/13 12:46] (current) – [Known bugs] gdangelo
Line 1: Line 1:
 ====== ARTÌS: Advanced RTI System ====== ====== ARTÌS: Advanced RTI System ======
  
-The Advanced RTI System (ARTÌS) is a middleware for Parallel and Distributed Simulation (PADS) supporting massively populated models [[http://gdangelo.web.cs.unibo.it/pool/papers/gdangelo-PERF-2005.pdf|[FIRB2005]]]. The design of the middleware is partially inspired by the IEEE 1516 standard but new specific features have been introduced to improve scalability and simulation speed.+The Advanced RTI System (ARTÌS) is a middleware for Parallel and Distributed Simulation (PADS) supporting massively populated models [[http://gdangelo.web.cs.unibo.it/pool/papers/gdangelo-PERF-2005.pdf|[FIRB2005]]]. The design of the middleware is partially inspired by the IEEE 1516 standard but new specific features have been introduced to improve scalability and execution speed.
  
-A PADS is composed by many interacting Logical Processes (LPs), therefore the execution speed is highly affected by the communication performance between the LPs. The approach followed by ARTÌS is adaptive and exploits the LPs physical allocation. Specifically, a couple of LPs running on the same physical execution unit (e.g. a multiprocessor PC) will communicate using the low latency shared memory. In the case of LPs connected by LAN, WAN or Internet, they will communicate using the R-UDP/IP or TCP/IP protocols. The protocol choice is adaptive and based on the network performances. +A PADS is composed of many interacting Logical Processes (LPs), therefore the execution speed is highly affected by the communication performance between the LPs. The approach followed by ARTÌS is adaptive and exploits the LPs physical allocation. Specifically, a couple of LPs running on the same physical execution unit (e.g. a multiprocessor PC) will communicate using the low latency shared memory. In the case of LPs connected by LAN, WAN or Internet, they will communicate using the R-UDP/IP or TCP/IP protocols. The protocol choice is adaptive and based on the network performances.
  
 Other important features of the middleware are the multi-threading support, the implementation of data structures specifically tailored for the event-list management, the capability to detect and adapt to Hyper-Treading and Multi-Core CPUs, and the support for the Simulation Cloning and Concurrent Replication Other important features of the middleware are the multi-threading support, the implementation of data structures specifically tailored for the event-list management, the capability to detect and adapt to Hyper-Treading and Multi-Core CPUs, and the support for the Simulation Cloning and Concurrent Replication
 of Independent Runs [[http://gdangelo.web.cs.unibo.it/pool/papers/gdangelo-HPCC-2005.pdf|[HPC2005]]]. of Independent Runs [[http://gdangelo.web.cs.unibo.it/pool/papers/gdangelo-HPCC-2005.pdf|[HPC2005]]].
  
-On to of ARTÌS has been implemented a framework called GAIA that provides an easy to use programming environment for the implementation of simulations. The framework follows the Multi-Agent-System (MAS) paradigm and introduces many enhancements for the reduction of the communication overhead and for load-balancing the execution architecture.+On top of ARTÌS has been implemented a framework called GAIA/GAIA+ that provides an easy to use programming environment for the implementation of simulations. The framework follows the Multi-Agent-System (MAS) paradigm and introduces many enhancements for the reduction of the communication overhead and for load-balancing the execution architecture.
  
 +===== Features =====
 +
 +  * Discrete Event Simulation (DES)
 +  * Parallel Discrete Event Simulation (PDES)
 +  * Sequential (monolithic) simulation
 +  * Parallel and Distributed Simulation (PADS)
 +  * Support for TCP/IP and shared memory communication
 +  * Conservative, optimistic and time-stepped synchronization
 +  * Multi-core / Hyper-Threading support
 +  * Source code is partially available
 +  * GNU/Linux 32 and 64 bit support
 +  * C and Java bindings
 +
 +===== Download =====
 +
 +**The last version of ARTÌS can be freely downloaded from the [[pads:download|Download]] page.**
 +
 +===== Documentation =====
 +
 +**For more information on the ARTÌS installation and usage please see the [[pads:artisinst|ARTÌS & GAIA HOWTO]]. **
 +
 +===== Citation =====
 +
 +To cite the ARTÌS software use:
 +
 +<file text gda-perf-2005.txt>
 +Luciano Bononi, Michele Bracuto, Gabriele D'Angelo, Lorenzo Donatiello.
 +Scalable and Efficient Parallel and Distributed Simulation of Complex, Dynamic and Mobile Systems. 
 +Proc. of the IEEE FIRB-Perf Workshop on Techniques Methodologies and Tools for Performance Evaluation of Complex Systems.
 +Torino, Italy, September, 2005. ISBN 0-7695-2447-8 
 +</file>
 +
 +
 +If you use BibTeX for LaTeX, use:
 +
 +<file latex gda-perf-2005.tex>
 +@inproceedings{gda-perf-2005,
 +  author = {Bononi, Luciano and Bracuto, Michele and D'Angelo, Gabriele and Donatiello, Lorenzo},
 +  title = {Scalable and Efficient Parallel and Distributed Simulation of Complex, Dynamic and Mobile Systems},
 +  booktitle = {Proceedings of the 2005 Workshop on Techniques, Methodologies and Tools for Performance Evaluation of Complex Systems},
 +  year = {2005},
 +  isbn = {0-7695-2447-8},
 +  url = {http://portal.acm.org/citation.cfm?id=1114282.1114476},
 +  doi = {10.1109/FIRB-PERF.2005.17},
 +  publisher = {IEEE Computer Society},
 +  address = {Washington, DC, USA},
 +}
 +</file>
 +  
 +===== Known bugs =====
 +
 +  * The MPI-based version of ARTÌS does not support other synchronization methods than Chandy-Misra-Bryant (CMB) and Time-Stepped (TS).
pads/artis.1309441323.txt.gz · Last modified: 2011/06/30 13:42 by gdangelo

Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki