Suporte a ambientes virtuais colaborativos de larga escala em redes peer-to-peer, com gerenciamento de distribuição de dados em conformidade com o padrão HLA

Abstract

In Large Scale Collaborative Virtual Environments LSCVEs, extensive synthetic 3D environments are shared among a large number of users that collaborate towards the same objective. As all users in these environments need immediate answer for their actions and these actions must be sent to all participating users, the application success depends not only on a strong graphic processing but also in the capacity of the network to deliver information in time. Data distribution management algorithms in conformity with the High Level Architecture / Run Time Infrastructure (HLA/RTI) pattern for parallel and distributed simulations have been used to reduce latency and to limit and control the data amount exchanged during simulations. The data distribution of LSCVEs is generally made by one of these communication models: client/server or Peer-to-peer. Differently of the client/server where the server can be a bottleneck of the network, in Peer-to-peer solutions the tasks are distributed and consequently applications are scalable, i.e., support a crescent client number. Motivated by these largely studied and commonly used technologies, a fault tolerant and low latency solution was searched, addressing the strict requirements of large scale collaborative virtual environments simulations. This was made in conformity with the HLA pattern, which users, using their own computer connected at Gnutella network, can participate in simulation sessions without the limitations found in kits that support distributed simulations like the RTI-Kit existent. For this reason an architecture was proposed with data distribution management in conformity with the HLA/RTI and that use the Gnutella Peer-to-peer communication model to make available and to share these environments over Mobile Ad-Hoc Networks (MANETs). Towards this propose simulations were made comparing the RTI-Kit developed by Georgia Tech with one speed objects and the RTI-Kit Adapted with varied speed objects in a cluster. The evaluation of the total time of the federation execution, the total number of the multicast messages generated and the total number of messages exchanged by the grid originated graphics that show up considerable increasing in the time and the number of messages exchanged by the grid mainly. In the same way the fault tolerant technique was evaluated.