Field-programmable gate arrays (FPGAs) are customizable digital circuits into which highly optimized data paths can be programmed to process large amounts of data at incredible rates, essentially creating power-efficient custom accelerators for any algorithm. The inherent programmability of FPGAs allows for reconfiguration within a fraction of a second. While FPGAs have traditionally been embedded into antennas, radars, network routers, and video broadcast equipment, they have recently become more prevalent in data centers for accelerating analytics, databases, image recognition, machine learning, search, and storage.
Michal Skiba explains how Intel is helping developers accelerate their cloud applications through a software stack that simplifies access to FPGAs from virtual machines and containers, enabling application developers to focus on their core algorithms without worrying about the intricacies and optimization of communication buses, management systems, and low-level drivers. The stack offers interfaces to orchestration frameworks for provisioning and managing FPGAs at scale, as well as numerous control points throughout to tune performance. Intel intends to enable an ecosystem of partners that would provide production-grade workload cores, libraries, integrated systems, and solutions, further simplifying the use of FPGAs.
This session is sponsored by Intel.
Michal Skiba is principal product line manager at Intel, where he leads product planning and marketing for Intel’s FPGA virtualization framework. Previously, he was a product line manager for Cisco’s data center switches. Michal holds an MASc from the University of Waterloo and a BASc from the University of Toronto. He has led several mountain-climbing expeditions in the Himalayas.
©2017, O'Reilly Media, Inc. • (800) 889-8969 or (707) 827-7019 • Monday-Friday 7:30am-5pm PT • All trademarks and registered trademarks appearing on oreilly.com are the property of their respective owners. • email@example.com