Turbocharging Artificial Intelligence With NGD Systems Computational Storage


Turbocharging Artificial Intelligence With NGD Systems Computational Storage

We have all seen demonstrations of the capabilities of Artificial Intelligence (AI)-based imaging applications, from facial recognition to computer vision assisted application platforms. However, scaling these imaging implementations to Petabyte-scale for real-time datasets is problematic because:

  1. Traditional databases contain structured tables of symbolic information that is built like a table with each image having a row in the index. The tables are then cross-linked and mapped to added indexes. For performance reasons, these index tables must be maintained in memory.
  2. Utilizing traditional implementations for managing datasets requires that data be moved from storage to server memory, after which the applications can analyze the data. Since these datasets are growing to Petabyte scale, the ability to analyze completedatasets in a single server’s memory set is all but impossible.

Utilizing brute-force methods on modern imaging databases, which can be in the petabyte-size range, is often both incredibly difficult and enormously expensive. This has forced organizations with extremely large image databases to look for new approaches to image similarity search and more generally to the problem of data storage and analysis.

To read the whole pager, please fill in below details.

[contact-form-7 404 "Not Found"]

About the author

vvdlinden administrator

Leave a Reply