Browsing by Author "Vega Caicedo, Christian Fernando"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
Item Explicit feedback for congestion control in Science DMZ cyberinfrastructures based on programable data-plane switches.(Universidad de Concepción, 2023) Vega Caicedo, Christian Fernando; Figueroa, Miguel; Pezoa, JorgeA massive amount of scientific data is generated daily in different areas of knowledge. The challenges of research are increasingly sophisticated and demand a higher amount of information to make decisions, and also technological tools are increasingly more available for researchers. In many cases, this information, in addition to being collected and processed, must be transmitted among different research centers located over considerable distances. Internet and other dedicated cyberinfrastructures are needed to achieve this goal. Transferring scientific information between remote sites involves challenges in performance, security, coexistence, and resource allocation. Therefore, Energy Science Network (ESnet) proposed the concept of Science DMZ that provides design patterns for an optimized network environment to exchange scientific data. Due to their nature, the Scientific Big Data (SBD) flows are connection-oriented, demanding high and constant throughput under low delay and jitter to reach reasonable transmission times. This imposes fundamental challenges to network design, especially congestion control mechanisms, buffer estimation, and anomaly detection, which are tackled in this thesis. This dissertation addresses two disruptive technologies to respond to these challenges: data-driven models and data-plane programmable devices. We evaluated our solutions through production networks and testbeds using a hardware-based data plane and routing devices. Results showed that the proposed solutions could effectively improve the networks’ performance and effectively adapt the Science DMZ design to non-dedicated networks.