Abstract: Techniques for determining insight are described. An exemplary method includes receiving a request to provide insight into potential abnormal behavior; receiving one or more of anomaly information and event information associated with the potential abnormal behavior; evaluating the received one or more of the anomaly information and event information associated with the abnormal behavior to determine there is insight as to what is causing the potential abnormal behavior and to add to an insight at least two of an indication of a metric involved in the abnormal behavior, a severity for the insight indication, an indication of a relevant event involved in the abnormal behavior, and a recommendation on how to cure the potential abnormal behavior; and providing an insight indication for the generated insight.

Abstract: Techniques for anomaly detection are described. An exemplary method includes receiving a request to detect anomalies, using an anomaly detection service, in time series data using one or more detectors; configuring the anomaly detection service by: generating a configuration for the anomaly detection service based on at least in part on one or more of the request the time series data, and metadata, wherein the configuration identifies at least one particular detector of the one or more detectors, and configuring the anomaly detection service using the generated configuration; evaluating the time series data for an anomaly using the configured anomaly detection service by: observing potentially anomalous behavior using the identified at least one particular detector of the one or more detectors, and generating an anomaly indication.

Abstract: Techniques of gracefully switching to direct connections for devices behind NAT devices are described. A server device behind a first NAT device receives a request to establish one or more peer-to-peer connections with a client device behind a second NAT device. The server device can establish a first peer-to-peer connection with the client device through a relay device that is publicly reachable by both the server device and the client device. The server device negotiates a second peer-to-peer connection with the client device, while the client device downloads content from the server device over the first peer-to-peer connection. The second peer-to-peer connection is on a network route that is different from a route of the first peer-to-peer connection. Once the second peer-to-peer connection is established, the server device and the client device can communicate with one another using the second peer-to-peer connection.

Abstract: Systems and methods for gracefully switching to direct connections for devices behind NAT devices are described. One of the methods includes receiving, by a server device logically located behind a first NAT device, a request to establish one or more peer-to-peer connections for providing content from the server device to a client device logically located behind a second NAT device. The method includes providing, by the server device to the client device, a control message for establishing a multiplexed connection under a multiplexed connection protocol, in which the control message is sent to the client device over a first peer-to-peer connection through a first network route including a relay device that is logically located between the first NAT device and the second NAT device. The method includes establishing the multiplexed connection between the client device and the server device, wherein the multiplexed connection includes a plurality of apparently separate and apparently independent connections.

Abstract: Methods, systems and apparatus, including computer programs encoded on computer storage media, for video content classification. A video content classification system classifies input video data frame by frame. The video content classification system performs customized discrete cosine transform (DCT) operations on each frame. The video content classification system then determines, based on results of the DCT operations, how much image information in the frame may be lost due to compression. If the loss is estimated to be below a threshold value, the video content classification system applies a high data compression ratio algorithm to compress the frame. Otherwise, the video content classification system applies a low data compression ratio algorithm to compress the frame. The video content classification system may generate output video data that is compressed differently frame by frame.

Abstract: Methods, systems and apparatus, including computer programs encoded on computer storage media, for video content classification. A video content classification system classifies input video data frame by frame. The video content classification system performs customized discrete cosine transform (DCT) operations on each frame. The video content classification system then determines, based on results of the DCT operations, how much image information in the frame may be lost due to compression. If the loss is estimated to be below a threshold value, the video content classification system applies a high data compression ratio algorithm to compress the frame. Otherwise, the video content classification system applies a low data compression ratio algorithm to compress the frame. The video content classification system may generate output video data that is compressed differently frame by frame.

Abstract: Techniques of hierarchical garbage collection in an object-relational database system are described. When the object-relational database system receives a request to delete an object, the object-relational database system marks the object for deletion. A garbage collection procedure executing in the background can perform iterations of sweep and mark to delete the object and children objects of the object. In each iteration, a layer or “generation” of objects is deleted. The garbage collection procedure can continue the iterations until the object and all descendants of the objects are deleted.

Abstract: Techniques of hierarchical garbage collection in an object-relational database system are described. When the object-relational database system receives a request to delete an object, the object-relational database system marks the object for deletion. A garbage collection procedure executing in the background can perform iterations of sweep and mark to delete the object and children objects of the object. In each iteration, a layer or “generation” of objects is deleted. The garbage collection procedure can continue the iterations until the object and all descendants of the objects are deleted.

Abstract: A method, system, and computer product for processing information embedded in a text file with a grammar programming language is provided. A text file is parsed according to a set of rules and candidate textual shapes corresponding to potential interpretations of the text file are provided by compiling a script. An output is provided, which may include either a processed value corresponding to a particular textual shape, or a textual representation of the text file that includes generic data structures that facilitate providing any of the candidate textual shapes, where the generic data structures are a function of the set of rules.