WIN307: Searching Inside Video at Petabyte Scale Using Spot - podcast episode cover

WIN307: Searching Inside Video at Petabyte Scale Using Spot

Dec 25, 201645 min
--:--
--:--
Listen in podcast apps:

Episode description

Video is a 'last-mile problem' for search technology. Unlike webpages, documents, and email, content in videos has traditionally been impossible to search. Recent advances in automated speech and text recognition, however, let businesses and universities search inside video assets as easily as inside textual content. In this session, you'll learn how Panopto is using AWS to solve the video-search problem at scale, while saving over 50% in operating costs by taking advantage of Spot instances. We discuss the cross-platform architecture that combines Windows and Linux to provide cost-effective video processing and search indexing. We also dive deep into scaling Spot elastically based on user demand, handling fallback situations when instances are revoked, and using the Spot bidding process to optimize cost structure. Finally, we discuss future plans to reduce operating costs even further through Spot fleets and grid processing.
WIN307: Searching Inside Video at Petabyte Scale Using Spot | AWS re:Invent 2016 podcast - Listen or read transcript on Metacast