Abstract: Techniques for partitioning data flow operations between execution on a compute instance and an attached accelerator instance are described. A set of operations supported by the accelerator is obtained. A set of operations associated with the data flow is obtained. An operation in the set of operations associated with the data flow is identified based on the set of operations supported by the accelerator. The accelerator executes the first operation.

Abstract: Systems and methods for Zero-day Data Loss Protection (DLP) having enhanced file upload processing are provided. One method may include capturing and sending file upload context (e.g. folder name, metadata, an active URL, etc.) associated with the scheduled file or folder upload to a DLP filesystem driver. For example, the method may include detecting whether a single/multi-file upload, a folder upload, or a drag-and-drop operation exists, through interception of the shell dialog API, the browse folder API, or the drop process interface, respectively. Further, the method may include generating a file upload cache including the file upload context, prior classification entries, and a timestamp indicating when the scheduled file or folder upload was last modified; such that, the DLP filesystem driver may intercept and process the file open call based upon the file upload cache. Accordingly, the file may be processed in accordance with a prior file classification, file/domain filter, or DLP policy.