Edge-Optimized Data Pipelines: Engineering for Low-Latency AI Processing
Keywords:
Edge Computing, Low Latency, Data Pipelines, AI Processing, Real-Time InferenceAbstract
As AI spreads throughout many industries, its use is gradually moving from more centralized cloud systems to edge environments, where decisions have to be made quickly & data is produced. Edge computing improves actual time responsiveness by allowing closeness of processing to the data source, hence lowering more reliance on cloud connectivity. Constructing data pipelines that are not just fast but also intended for low latency & lowest resource utilization will help to achieve more successful AI processing at the edge. The necessary relevance of building edge-optimized data pipelines is investigated in this work. We start by examining the several needs and constraints of edge computing—such as limited bandwidth, limited hardware capabilities & the need for actual time data processing—and their interaction with the rising deployment of AI in fields including autonomous vehicles, smart manufacturing & telemedicine. Emphasizing best methods in data intake, preprocessing, transformation & model inference, the paper also investigates the architecture of modern edge data pipelines. We investigate low-latency goals by means of model compression, stream processing & data prioritizing such that accuracy is maintained. With actual world case studies and technological insights, we show how businesses are solving more latency problems and guaranteeing consistent field performance. In the end, the article underlines that edge-optimized pipelines are not just a technical necessity but also a basic enabler for AI systems that have to run with instant operation. This article aims to give architects and engineers working at the junction of AI and edge computing a conceptual framework and pragmatic guidance.
Downloads
References
Owaida, Muhsen, et al. "Lowering the latency of data processing pipelines through FPGA based hardware acceleration." Proceedings of the VLDB Endowment 13.1 (2019): 71-85.
Diao, Yanlei, Abhishek Roy, and Toby Bloom. "Building Highly-Optimized, Low-Latency Pipelines for Genomic Data Analysis." CIDR. 2015.
De Prado, Miguel, et al. "Bonseyes AI Pipeline-bringing AI to you." End-to-end integration of data, algorithms and deployment tools, arxiv. org/abs/1901.05049 (2019).
Cheng, Yang, et al. "Dlbooster: Boosting end-to-end deep learning workflows with offloading data preprocessing pipelines." Proceedings of the 48th International Conference on Parallel Processing. 2019.
Possa, Paulo, et al. "P2ip: a novel low-latency programmable pipeline image processor." Microprocessors and Microsystems 39.7 (2015): 529-540.
Ali, Zafer, and Henrietta Nicola. "Accelerating Digital Transformation: Leveraging Enterprise Architecture and AI in Cloud-Driven DevOps and DataOps Frameworks." (2018).
Yasodhara Varma Rangineeni, and Manivannan Kothandaraman. “Automating and Scaling ML Workflows for Large Scale Machine Learning Models”. JOURNAL OF RECENT TRENDS IN COMPUTER SCIENCE AND ENGINEERING ( JRTCSE), vol. 6, no. 1, May 2018, pp. 28-41
Crankshaw, Daniel, et al. "Inferline: Ml inference pipeline composition framework." arXiv preprint arXiv:1812.01776 (2018).
Salehe, Mohammad, et al. "Videopipe: Building video stream processing pipelines at the edge." Proceedings of the 20th international middleware conference industrial track. 2019.
Fowers, Jeremy, et al. "Inside Project Brainwave's Cloud-Scale, Real-Time AI Processor." IEEE Micro 39.3 (2019): 20-28.
Atri, Preyaa. "Design and Implementation of High-Throughput Data Streams using Apache Kafka for Real-Time Data Pipelines." International Journal of Science and Research (IJSR) 7.11 (2018): 1988-1991.
Kumar, Tambi Varun. "REAL-TIME COMPLIANCE MONITORING IN BANKING OPERATIONS USING AI." (2018).
Isah, Haruna, et al. "A survey of distributed data stream processing frameworks." IEEE Access 7 (2019): 154300-154316.
Anusha Atluri, and Teja Puttamsetti. “The Future of HR Automation: How Oracle HCM Is Transforming Workforce Efficiency”. JOURNAL OF RECENT TRENDS IN COMPUTER SCIENCE AND ENGINEERING ( JRTCSE), vol. 7, no. 1, Mar. 2019, pp. 51–65
Khalid, Mumtaz, and Jonny Bairstow. "Next-Gen Enterprise Architecture: Harnessing AI, Cloud, DevOps, and DataOps for Scalability." (2019).
Chard, Ryan, et al. "DLHub: Model and data serving for science." 2019 IEEE International Parallel and Distributed Processing Symposium (IPDPS). IEEE, 2019.
Anusha Atluri. “The Revolutionizing Employee Experience: Leveraging Oracle HCM for Self-Service HR”. JOURNAL OF RECENT TRENDS IN COMPUTER SCIENCE AND ENGINEERING ( JRTCSE), vol. 7, no. 2, Dec. 2019, pp. 77-90
Guo, Xiangyu, et al. "GPU-acceleration on a low-latency binary-coalescence gravitational wave search pipeline." Computer Physics Communications 231 (2018): 62-71.
