Examinando por Autor "Leger, Paul"

Mostrando 1 - 3 de 3
  • Miniatura
    PublicaciónAcceso abierto
    Ad Hoc systems management and specification with distributed Petri Nets.
    (Universidad de los Andes, 2021) Sosa Flórez, Juan Sebastián; Cardozo Álvarez, Nicolás; Takahashi Rodríguez, Silvia; Leger, Paul
    Managing mobile ad hoc systems is a difficult task due to the high volatility of the systems¿ topology. Ad hoc systems are commonly defined by means of their constituent entities and the relationships between such entities, however, a formal specification and run-time execution model is missing. The benefit of a formal specification is that it can enable reasoning about local and global system properties, for example, determining whether the system can reach a given state. We propose a Petri net-based specification and execution model to manage ad hoc distributed systems. Our model enables spontaneous communication between previously unknown system components. The model is locally equivalent to standard Petri nets, and hence could be used for the verification of properties for system snapshots static with respect to connections and disconnection, in which it is possible to analyze liveness, reachability, or conflicts. We validate the usability of our distributed ad hoc Petri net model by modeling distributable systems as described by existing distributed Petri nets approaches. Additionally, we demonstrate the ap- plicability and usability of the proposed model in distributed ad hoc networks by implementing the communication behavior of two prototypical ad hoc network applications, disaster and crisis management, and VANETs, successfully validating the appropriate behavior of the system in each case.
  • Miniatura
    PublicaciónAcceso abierto
    Analysis of WebRTC signaling
    (Universidad de los Andes, 2021-12-13) Rodríguez Baquero, Diego; Linares Vasquez, Mario; 80728886; Leger, Paul; Cardozo Álvarez, Nicolás; The Software Design Lab
    Web Real-Time-Communications (WebRTC) is a modern open framework that allows peer-to-peer (P2P) communication. It is currently the only mechanism that a web browser has to perform P2P connections; mostly used for video and audio calls, but can be used to communicate any arbitrary data. Other popular use cases include: P2P chat, real time videogames and file sharing. As on any P2P system, a mechanism must exist that allows bootstrapping the P2P connection (making the two peers aware of their address:port, NAT hole-punching, etc.). Two examples are bittorrent trackers, and the bootstrap nodes hard-coded in Bitcoin's full-node software. In the WebRTC framework, this process is called signaling, it allows one peer's connection offer to be shared with a responding second peer, which then replies with a connection answer that must be relayed back to the first peer. This signaling must be performed through a bi-directional transport (e.g. polling, websockets). In this thesis we propose Multti, a new no-delay transport similar to WebSockets that is both less resource consuming and more scalable. Multti uses the browser Fetch API to send data from the client, and uses an EventSource stream to receive Server-Sent Events (SSE) immediately. Harnessing the reusability of HTTP/2 sockets, a browser may open multiple streams and perform multiple HTTP POSTs using only one socket. Multiple experiments were performed with varying configurations such as concurrency, number of peers, and number of browser processes while measuring server's memory consumption and TCP sockets. Results show that the proposal consumes both less memory and less sockets in most scenarios. While using HTTP/1 is possible with the proposed implementation, the benefits would be lost. To conclude, a new client and server WebRTC signaling implementation is proposed using Fetch and SSE on top of HTTP/2, that consumes less resources and is more scalable receiving data. Future work on Multti would focus on making it multi-threaded and generic, to allow not only WebRTC signaling, but any bi-directional communication to happen.
  • Miniatura
    PublicaciónAcceso abierto
    Points-to analysis for context-oriented JavaScript programs
    (Universidad de los Andes, 2022-06-15) Cárdenas Landazábal, Sergio Eduardo; Cardozo Álvarez, Nicolás; 80870736; Takahashi Rodriguez, Silvia; Leger, Paul; FLAG
    The purpose of this work is to improve the results of analyzers that could use the points-to sets as an input such as type checking by improving the precision of a points-to analysis for context-oriented JavaScript programs. We implement points-to analysis for context-oriented JavaScript programs, that consider both dynamic capabilities of the language and functionalities of COP. To narrow down the problem, we implement the analyzer for a specific implementation of COP for JavaScript.