When it comes to machine learning, knowing how to get the most out of our computing power is essential for building efficient systems and useful products for market. Here at Aspire we’re proud to have an abundance of inventive minds and pioneering experts in optimization that know how to do just that.
The patent would allow RS-DVR providers to drastically reduce storage for video content with a new file system that lets individual files share any common elements. In the case of Remote Storage DVR’s, where any given program requires thousands and thousands of copies for each individual user, Rogers’ new system could save vast amounts of storage.
When Cablevision introduced RS-DVR—a DVR service that stores the content in the cloud rather than on an expensive, bulky home device—a consortium of copyright holders sued Cablevision for unlicensed rebroadcasting. Cablevision eventually won the case, but with a caveat: in order for the RS-DVR service to not be a rebroadcast, there must be individual copies of recorded programs for each subscriber. And that has been the law of the land ever since.
With traditional files systems, a single file has its own unique set of bytes, sectors, and blocks for each file, which means that providers like Cablevision had to store thousands of large files for a single recorded program for each subscriber. Rogers’ invention is a new file system that maintains individual files, but the mapping from those files would share sectors and blocks that are the same, so that the raw contents of the file on disk is stored only once to be shared, not duplicated.
“Thus, the legal requirement of separate files for recording is maintained” Rogers says, “but the storage at the disk level is optimized.”
Rogers’ expertise in optimization and inventiveness in systems architecture has made him an essential leader to our core technology department, as they develop machine learning solutions to some of the world’s most pressing problems. Although the insights that machine learning produces can be invaluable, it can also take massive amounts of computing power, and that comes at a cost. At Aspire, Rogers is building on his past success to optimize our processes and predictive models, so we can harness those important machine learning insights with less computing power.