Engineering professors will use Purdue's real-world testbeds for edge computing in collaboration with Cisco
Two faculty members in the Elmore Family School of Electrical and Computer Engineering recently were awarded a contract from Cisco Research to create a new dimension of edge computing that incorporates data analytics for IoT applications using programmable network switches.
Bagchi’s contributions will include his current work, an accepted paper at Eurosys 2022 that demonstrated how to build adaptive video object detection systems on mobile GPUs, as well as a paper published in CVPR 2022 that explained how such a video system could be made lightweight enough to provide real-time processing.
“Our work will enable edge computing for two popular and emerging applications — real-time analytics from augmented reality (AR) cameras and sensing and actuation for controlling operations on a manufacturing pipeline. We have real-world testbeds at Purdue for both these application domains, which served as a crucial differentiator in the proposal process,” Bagchi said.
The proposal, titled “Accelerating Edge Computing Using In-Network Computing,” also will rely on Shrivastav’s research, which is outlined in a paper in SIGCOMM 2022 that introduced stateful processing on multi-Tbps (terabits per second) programmable switches comprising multiple parallel processing pipelines.
“We will achieve this unprecedented capability of high-rate streaming analytics by employing the concept of in-network computing. This is the idea that part of the data analytics will happen inline in programmable network switches, rather than only on end-point hosts,” Shrivastav said.
Such programmable network switches were pioneered by Barefoot Networks in 2016 and are now commercially manufactured by several big switch vendors, including Cisco, Broadcom and Juniper.
“The mission of Cisco Research is to conduct and foster cutting-edge research in areas of strategic interest to Cisco," said Ramana Kompella, head of Cisco Research. "As part of this mission, we partner with several top universities with broad research strengths to fund and work with faculty on interesting projects that can make a material impact to Cisco’s future innovations. We are super excited to continue our strategic partnership with Purdue University through this new set of awards. Working with professors Shrivastav and Bagchi adds a unique element to our research portfolio, programmable networks for edge analytics.”
Both Bagchi and Shrivastav are members of Purdue’s Center for Resilient Infrastructures, Systems and Processes (CRISP), which Bagchi leads.
The Cisco researchers for the project are Myungjin Lee, Ashish Kundu and Jayanth Srinivasa.
“More data is generated at the edge, and gaining high-quality insights fast near data sources can bring significant benefits by reducing data shipping and hosting costs at the cloud,” according to a statement from the Cisco team. “We aim to develop new technologies for fast and efficient edge data analytics for security and machine learning. The collaboration with the Purdue research team will pave the way for achieving our goals.”