Berkeley eXtensible Processing Engine (BXPE)

Featured

An FPGA-based edge network processing pipeline for real-time scientific data processing at 100GbE link rates, developed in collaboration with ESnet.

FPGA100GbEEdge ComputingReal-time ProcessingCo-designESnet
2020–2022

Overview

The Berkeley eXtensible Processing Engine (BXPE) is an FPGA-based edge network processing pipeline designed for real-time data processing using hardware/software co-design principles. Developed in collaboration with ESnet at Lawrence Berkeley National Lab, BXPE enables scientific instruments to process data at 100 Gigabit Ethernet link rates.

Key Features

  • 100GbE Processing: Ingest and process data from scientific instruments at full 100 Gigabit Ethernet line rate
  • Co-design Approach: Hardware/software co-design for optimal performance and flexibility
  • DSP Integration: Leverages FPGA DSP engines for compute-intensive operations
  • Real-time Analysis: Enables on-the-fly data reduction and analysis at the network edge

Applications

National Center for Electron Microscopy (NCEM)

  • Implemented Center-of-Mass computation using FPGA DSP engines
  • Real-time processing of electron microscopy data streams

Advanced Light Source (ALS)

  • Implemented Convolution computation for beamline data
  • High-throughput image processing at detector speeds

Collaboration

BXPE was developed in collaboration with ESnet (Energy Sciences Network), working directly with scientists at NCEM and ALS to address their real-time data processing needs.

Impact

BXPE demonstrates the potential of FPGA-based edge computing for scientific applications, enabling new experimental capabilities that require immediate data feedback and reduction at unprecedented data rates.