CC* Integration-Small: Network-Aware Edge Computing for Real-time Wildfire Detection

物联网(IoT)设备的激增促进了各种科学应用的发展，从智慧城市倡议到环境危害监测系统。在大多数这样的应用中，快速处理物联网传感器捕获的数据对于及时预防自然灾害至关重要。虽然传统的基于云的数据处理管道提供了具有成本效益和性能效率的解决方案，但要满足低延迟和高带宽等危险监控系统的严格性能要求往往是具有挑战性的。边缘计算作为一种引人注目的解决方案出现，通过使计算处理更接近数据源来弥合这一差距。该项目开发了一个边缘计算框架，以满足时间敏感的分布式科学应用程序(如野火监测)的计算需求。边缘计算框架有可能惠及野火监测以外的其他领域，如自动驾驶车辆和应急响应系统。该项目开发了一个边缘计算框架，通过将高精度系统监控与全面的应用程序评测相结合来优化任务调度问题。它开发了一个可扩展的资源监控系统，使用轻量级监控代理和P4可编程网络设备来监控计算资源(如边缘服务器和云实例)和网络资源的状态。此外，该项目还进行应用评测，以提取跨各种边缘服务器和云实例配置的资源利用率和任务执行时间的基本指标。将调度问题描述为多目标优化问题，探索了混合整数线性规划、遗传算法和启发式方法等多种优化方法。最后，该团队将野火检测项目(AlertWildfire)作为一个用例，以演示所提议的框架的有效性。该项目由高级网络基础设施办公室、已建立的激励竞争研究计划(EPSCoR)和计算机和网络系统部门联合资助。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The proliferation of Internet of Things (IoT) devices has facilitated the development of various scientific applications, from smart city initiatives to environmental hazard monitoring systems. In most of these applications, swift processing of data captured by IoT sensors is crucial to prevent natural disasters in a timely manner. While conventional cloud-based data processing pipelines offer cost-effective and performance-efficient solutions, it is often challenging to meet stringent performance requirements of hazard monitoring systems such as low latency and high bandwidth. Edge computing emerges as a compelling solution to bridge this gap by bringing computational processing closer to the data source. This project develops an edge computing framework tailored to address the computational requirements of time-sensitive distributed scientific applications such as wildfire monitoring. The edge computing framework has the potential to benefit other domains beyond wildfire monitoring, such as autonomous vehicles and emergency response systems.The project develops an edge computing framework that optimizes the task scheduling problem by combining high precision system monitoring with comprehensive application profiling. It develops a scalable resource monitoring system to monitor the status of compute resources (e.g., edge servers and cloud instances) and network resources using lightweight monitoring agents and P4 programmable network devices. Additionally, the project conducts application profiling to extract essential metrics regarding resource utilization and execution time of tasks across various edge server and cloud instance configurations. The scheduling is formulated as a multi-objective optimization problem, and various optimization methods such as mixed-integer linear programming, genetic algorithms, and heuristic methods are explored. Finally, the team targets a wildfire detection project (AlertWildfire) as a use case to demonstrate the effectiveness of the proposed framework. This project is jointly funded by Office of Advanced Cyberinfrastructure, the Established Program to Stimulate Competitive Research (EPSCoR), and the Division of Computer and Network Systems.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.