User Tools

Site Tools


gossip-paper

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
Last revisionBoth sides next revision
gossip-paper [2016/05/02 07:11] – [Status] gdangelogossip-paper [2020/01/17 08:23] gda
Line 8: Line 8:
 ====== Abstract ====== ====== Abstract ======
  
-//This paper presents a study on data dissemination in unstructured Peer-to-Peer (P2P) network overlays. The absence of a structure in unstructured overlays eases the network management, at the cost of non-optimal mechanisms to spread messages in the network. Thus, dissemination schemes must be employed that allow covering a large portion of the network with a high probability (e.g.~gossip based approaches).  +//This paper presents a study on data dissemination in unstructured Peer-to-Peer (P2P) network overlays. The absence of a structure in unstructured overlays eases the network management, at the cost of non-optimal mechanisms to spread messages in the network. Thus, dissemination schemes must be employed that allow covering a large portion of the network with a high probability (e.g.~gossip based approaches). We identify principal metrics, provide a theoretical model and perform the assessment evaluation using a high performance simulator that is based on a parallel and distributed architecture. A main point of this study is that our simulation model considers implementation technical details, such as the use of caching and Time To Live (TTL) in message dissemination, that are usually neglected in simulations, due to the additional overhead they cause. Outcomes confirm that these technical details have an important influence on the performance of dissemination schemes and that the studied schemes are quite effective to spread information in P2P overlay networks, whatever their topology. Moreover, the practical usage of such dissemination mechanisms requires a fine tuning of many parameters, the choice between different network topologies and the assessment of behaviors such as free riding. All this can be done only using efficient simulation tools to support both the network design phase and, in some cases, at runtime.//
-We identify principal metrics, provide a theoretical model and perform the assessment evaluation using a high performance simulator that is based on a parallel and distributed architecture. A main point of this study is that our simulation model considers implementation technical details, such as the use of caching and Time To Live (TTL) in message dissemination, that are usually neglected in simulations, due to the additional overhead they cause. Outcomes confirm that these technical details have an important influence on the performance of dissemination schemes and that the studied schemes are quite effective to spread information in P2P overlay networks, whatever their topology. Moreover, the practical usage of such dissemination mechanisms requires a fine tuning of many parameters, the choice between different network topologies and the assessment of behaviors such as free riding. All this can be done only using efficient simulation tools to support both the network design phase and, in some cases, at runtime.//+
  
 ====== Keywords ====== ====== Keywords ======
Line 17: Line 16:
 ====== Status ====== ====== Status ======
  
-Submitted for publication. A draft version will be made available in the next weeks.+   * Gabriele D’Angelo, Stefano Ferretti, Highly intensive data dissemination in complex networks, Journal of Parallel and Distributed Computing, Volume 99, January 2017, Pages 28-50, ISSN 0743-7315, [[http://dx.doi.org/10.1016/j.jpdc.2016.08.004|http://dx.doi.org/10.1016/j.jpdc.2016.08.004]]
  
 ====== Resources ====== ====== Resources ======
  
 +  * Draft version of the paper available on [[https://arxiv.org/abs/1507.08417|arxiv]]
   * The LUNES source code is part of the [[pads:download|ARTÌS software distribution]]   * The LUNES source code is part of the [[pads:download|ARTÌS software distribution]]
   * Raw data obtained by the experiments, graph corpuses, processed data and scripts used to produce the figures: {{pads:LUNES-gossip-extra-r1.tar.bz2|download}}   * Raw data obtained by the experiments, graph corpuses, processed data and scripts used to produce the figures: {{pads:LUNES-gossip-extra-r1.tar.bz2|download}}
gossip-paper.txt · Last modified: 2020/01/17 10:33 by gda

Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki