Browsing by Author "Sanabria Quispe, Pablo"

Now showing 1 - 9 of 9
  • No Thumbnail Available
    Item
    A performance comparison of heuristics for scheduling jobs in hybrid mobile topologies
    (IEEE, 2019) Sanabria Quispe, Pablo; Tapia, T. F.; Benedetto Causa, José Ignacio; Hirsch, M.; Mateos, C.; Zunino, A.
    Hardware improvements of mobile devices have allowed them to be considered as first-class resources at the edge, which has led to the need to propose new techniques and scheduling heuristics to make efficient usage of these devices. Current scheduling heuristics available in the literature do not take into account mobile grid nodes that do not rely on batteries to work, a factor that is of great importance in the advent of mobile-edge computing. In this paper we propose a simple algorithm that incorporates a more advanced awareness of a mobile device's power supply and we show that incorporating this characteristic can enable these heuristics to manage edge resources in a more efficient way.
  • Thumbnail Image
    Item
    An Empirical Study of Mobile Code Offloading in Unpredictable Environments
    (2023) Sanabria Quispe, Pablo; Neyem, Andres; Sandoval Alcocer, Juan Pablo; Fernandez, Blanco Alison
    Mobile code offloading is a well-known technique for enhancing the capabilities of mobile platforms by transparently leveraging the resources to the cloud. Although this technique has been studied for years, little empirical evidence exists to demonstrate its alleged benefits in terms of performance in real-life situations. All studies conducted on this topic have so far been relegated to controlled environments in laboratory settings. As such, there is no evidence of how and how well this technique performs in real-life scenarios, where network unreliability is the norm. In this work, we present the first empirical study of an Android mobile application integrated with a code offloading framework being tested in the wild. We distributed an application that contains a set of benchmarks in APK format and deployed it on a wide gamut of Android devices to which we had no physical access. We carefully detail the methodology and infrastructure we used to monitor the benchmarks’ performance of 18 devices. Overall, our results show that the accuracy of the decision-making engine is heavily affected by a couple of factors, mainly the network diagnosis and connection type. Therefore, determining whether or not it is more convenient to execute a given task in the cloud is a difficult task. We summarize five lessons we learned by performing our experiment that we believe should be considered for future experiments in this area.
  • Thumbnail Image
    Item
    Code Offloading Solutions for Audio Processing in Mobile Healthcare Applications: A Case Study
    (IEEE, 2018) Sanabria Quispe, Pablo; Benedetto Causa, José Ignacio; Neyem, Andrés; Navón Cohen, Jaime; Poellabauer, C.
    In this paper, we present a real-life case study of a mobile healthcare application that leverages code offloading techniques to accelerate the execution of a complex deep neural network algorithm for analyzing audio samples. Resource-intensive machine learning tasks take a significant time to complete on high-end devices, while lower-end devices may outright crash when attempting to run them. In our experiments, offloading granted the former a 3.6x performance improvement, and up to 80% reduction in energy consumption; while the latter gained the capability of running a process they originally could not.
  • Thumbnail Image
    Item
    Connection-Aware Heuristics for Scheduling and Distributing Jobs under Dynamic Dew Computing Environments
    (2024) Sanabria Quispe, Pablo; Montoya Tapia, Sebastián Ignacio; Neyem, Andrés; Toro Icarte, Rodrigo Andrés; Hirsch, Matías; Mateos, Cristian
    Due to the widespread use of mobile and IoT devices, coupled with their continually expanding processing capabilities, dew computing environments have become a significant focus for researchers. These environments enable resource-constrained devices to contribute computing power to a local network. One major challenge within these environments revolves around task scheduling, specifically determining the optimal distribution of jobs across the available devices in the network. This challenge becomes particularly pronounced in dynamic environments where network conditions constantly change. This work proposes integrating the “reliability” concept into cutting-edge human-design job distribution heuristics named ReleSEAS and RelBPA as a means of adapting to dynamic and ever-changing network conditions caused by nodes’ mobility. Additionally, we introduce a reinforcement learning (RL) approach, embedding both the notion of reliability and real-time network status into the RL agent. Our research rigorously contrasts our proposed algorithms’ throughput and job completion rates with their predecessors. Simulated results reveal a marked improvement in overall throughput, with our algorithms potentially boosting the environment’s performance. They also show a significant enhancement in job completion within dynamic environments compared to baseline findings. Moreover, when RL is applied, it surpasses the job completion rate of human-designed heuristics. Our study emphasizes the advantages of embedding inherent network characteristics into job distribution algorithms for dew computing. Such incorporation gives them a profound understanding of the network’s diverse resources. Consequently, this insight enables the algorithms to manage resources more adeptly and effectively.
  • Thumbnail Image
    Item
    Enriching Capstone Project-Based Learning Experiences Using a Crowdsourcing Recommender Engine
    (IEEE, 2017) Diaz-Mosquera, Juan; Sanabria Quispe, Pablo; Neyem, Andrés; Parra Santander, Denis; Navón Cohen, Jaime
    Capstone project-based learning courses generate a suitable space where students can put into action knowledge specific to an area. In the case of Software Engineering (SE), students must apply knowledge at the level of Analysis, Design, Development, Implementation and Management of Software Projects. There is a large number of supportive resources for SE that one can find on the web, however, information overload ends up saturating the students who wish to find resources more accurate depending on their needs. This is why we propose a crowdsourcing recommender engine as part of an educational software platform. This engine based its recommendations on content from StackExchange posts using the project's profile in which a student is currently working. To generate the project's profile, our engine takes advantage of the information stored by students in the aforementioned platform. Content-based algorithms based on Okapi BM25 and Latent Dirichlet Allocation (LDA) are used to provide suitable recommendations. The evaluation of the engine was held with students from the capstone course in SE of the University Catholic of Chile. Results show that Cosine similarity over traditional bag-of-words TF-IDF content vectors yield interesting results, but they are outperformed by the integration of BM25 with LDA.
  • Thumbnail Image
    Item
    MobiCOP : A Scalable and Reliable Mobile Code Offloading Solution
    (2018) Benedetto Causa, José Ignacio; Valenzuela, Guillermo; Sanabria Quispe, Pablo; Neyem, Andrés; Navón Cohen, Jaime; Poellabauer, Christian
  • Thumbnail Image
    Item
    Solving Task Scheduling Problems in Dew Computing via Deep Reinforcement Learning
    (2022) Sanabria Quispe, Pablo; Tapia, Tomás Felipe; Toro Icarte, Rodrigo; Neyem, Andres
    Due to mobile and IoT devices’ ubiquity and their ever-growing processing potential, Dew computing environments have been emerging topics for researchers. These environments allow resource-constrained devices to contribute computing power to others in a local network. One major challenge in these environments is task scheduling: that is, how to distribute jobs across devices available in the network. In this paper, we propose to distribute jobs in Dew environments using artificial intelligence (AI). Specifically, we show that an AI agent, known as Proximal Policy Optimization (PPO), can learn to distribute jobs in a simulated Dew environment better than existing methods—even when tested over job sequences that are five times longer than the sequences used during the training. We found that using our technique, we can gain up to 77% in performance compared with using human-designed heuristics.
  • Thumbnail Image
    Item
    Towards a practical framework for code offloading in the Internet of Things
    (2019) Benedetto Causa, José Ignacio; González Cos, Luis Armando; Sanabria Quispe, Pablo; Neyem, Andrés; Navón Cohen, Jaime
  • Thumbnail Image
    Item
    Towards Native Code Offloading Platforms for Image Processing in Mobile Applications: A Case Study
    (IEEE, 2017) Valenzuela Gallegos, Guillermo Esteban; Neyem, Hugo Andres; Benedetto Causa, José Ignacio; Navon Cohen, Jaime; Sanabria Quispe, Pablo; Karmy, Juan A.; Balbontin, Felipe
    In this paper, we present a real-life case study to show the advantages of a new code offloading solution focused on improving both performance and energy consumption for image processing mobile applications. From our experiments, we found that offloading an image processing task would allow up to 5.7x speedup and 85% of reduction in energy consumption for low-end devices, and 1.7x speedup and 64.3% of reduction in energy consumption for high-end devices.