Abstract: A system for utilizing models derived from offline historical data in online applications is provided. The system includes a processor and a memory storing machine-readable instructions for determining a set of contexts of the usage data, and for each of the contexts within the set of contexts, collecting service data from services supporting the media service and storing that service data in a database. The system performing an offline testing process by fetching service data for a defined context from the database, generating a first set of feature vectors based on the fetched service data, and providing the first set to a machine-learning module. The system performs an online testing process by fetching active service data from the services supporting the media streaming service, generating a second set of feature vectors based on the fetched active service data, and providing the second set to the machine-learning module.

Abstract: A system for utilizing models derived from offline historical data in online applications is provided. The system includes a processor and a memory storing machine-readable instructions for determining a set of contexts of the usage data, and for each of the contexts within the set of contexts, collecting service data from services supporting the media service and storing that service data in a database. The system performing an offline testing process by fetching service data for a defined context from the database, generating a first set of feature vectors based on the fetched service data, and providing the first set to a machine-learning module. The system performs an online testing process by fetching active service data from the services supporting the media streaming service, generating a second set of feature vectors based on the fetched active service data, and providing the second set to the machine-learning module.

Abstract: A system for utilizing models derived from offline historical data in online applications is provided. The system includes a processor and a memory storing machine-readable instructions for determining a set of contexts of the usage data, and for each of the contexts within the set of contexts, collecting service data from services supporting the media service and storing that service data in a database. The system performing an offline testing process by fetching service data for a defined context from the database, generating a first set of feature vectors based on the fetched service data, and providing the first set to a machine-learning module. The system performs an online testing process by fetching active service data from the services supporting the media streaming service, generating a second set of feature vectors based on the fetched active service data, and providing the second set to the machine-learning module.

Abstract: Generating streaming decisions at one or more remote servers on behalf of a local client is disclosed. An indication of a content request, multi-dimensional client information, and local state information are received from a client device. A record specific to the client device is generated, in a data store, using the received local state information. The data store includes a plurality of records corresponding to respective different client devices. A set of instructions usable by the client device to obtain and play content is determined based at least in part the local state information received from the client device. The determined set of instructions is transmitted to the client device.

Abstract: Directing a content player to a content source is disclosed. A message from a content management system including information associated with a client that is currently requesting content is received. Based at least in part on the information included in the message, a set of content sources is determined. Instructions indicating the set of content sources for the client are sent to the content management system.

Abstract: Selecting an optimal bitrate range is disclosed. A request for content from a first client is received. An optimal subset of bitrates for the first client is determined. The optimal subset of bitrates for the first client is determined based at least in part on one or more real-time quality measures. A different optimal subset of bitrates is determined for a second client. An indication of the optimal subset of bitrates determined for the first client is provided to the first client. The first client is configured to obtain content according to the indication of the optimal subset of bitrates.

Abstract: Data collection management is disclosed. A data collection configuration is obtained. The data collection configuration is translated into executable code in a language usable to collect data. Data is collected using the executable code. The collected data values are provided as output. Metrics management is also disclosed. A configuration of a metric is obtained. The metric configuration includes a definition of how computation of the metric is to be performed and a mapping between a computation input and collected data. Collected data values are obtained based at least in part on the mapping. Metric values are computed according to the definition. One or more results associated with the computed metric values are stored.

Abstract: Custom traffic tagging on a control plane backend is disclosed. Information associated with a video session is obtained. Based at least in part on at least some of the obtained information associated with the video session, additional metadata to associate with the video session is determined. The additional metadata is determined from metadata obtained from one or more disparate sources. The video session is associated with the determined additional metadata.

Abstract: A system for utilizing models derived from offline historical data in online applications is provided. The system includes a processor and a memory storing machine-readable instructions for determining a set of contexts of the usage data, and for each of the contexts within the set of contexts, collecting service data from services supporting the media service and storing that service data in a database. The system performing an offline testing process by fetching service data for a defined context from the database, generating a first set of feature vectors based on the fetched service data, and providing the first set to a machine-learning module. The system performs an online testing process by fetching active service data from the services supporting the media streaming service, generating a second set of feature vectors based on the fetched active service data, and providing the second set to the machine-learning module.

Abstract: A system for utilizing models derived from offline historical data in online applications is provided. The system includes a processor and a memory storing machine-readable instructions for determining a set of contexts of the usage data, and for each of the contexts within the set of contexts, collecting service data from services supporting the media service and storing that service data in a database. The system performing an offline testing process by fetching service data for a defined context from the database, generating a first set of feature vectors based on the fetched service data, and providing the first set to a machine-learning module. The system performs an online testing process by fetching active service data from the services supporting the media streaming service, generating a second set of feature vectors based on the fetched active service data, and providing the second set to the machine-learning module.

Abstract: Generating streaming decisions at one or more remote servers on behalf of a local client is disclosed. An indication of a content request, multi-dimensional client information, and local state information are received from a client device. A record specific to the client device is generated, in a data store, using the received local state information. The data store includes a plurality of records corresponding to respective different client devices. A set of instructions usable by the client device to obtain and play content is determined based at least in part the local state information received from the client device. The determined set of instructions is transmitted to the client device.

Abstract: Directing a content player to a content source is disclosed. A message from a content management system including information associated with a client that is currently requesting content is received. Based at least in part on the information included in the message, a set of content sources is determined. Instructions indicating the set of content sources for the client are sent to the content management system.

Abstract: Selecting an optimal bitrate range is disclosed. A request for content from a first client is received. An optimal subset of bitrates for the first client is determined. The optimal subset of bitrates for the first client is determined based at least in part on one or more real-time quality measures. A different optimal subset of bitrates is determined for a second client. An indication of the optimal subset of bitrates determined for the first client is provided to the first client. The first client is configured to obtain content according to the indication of the optimal subset of bitrates.

Abstract: Controlling delivery of a video asset to a client is disclosed. A request for a manifest file is received. Manifest file components are retrieved. A dynamic manifest file is assembled from a set of manifest file components.

Abstract: Custom traffic tagging on a control plane backend is disclosed. Information associated with a video session is obtained. Based at least in part on at least some of the obtained information associated with the video session, additional metadata to associate with the video session is determined. The additional metadata is determined from metadata obtained from one or more disparate sources. The video session is associated with the determined additional metadata.

Abstract: Data collection management is disclosed. A data collection configuration is obtained. The data collection configuration is translated into executable code in a language usable to collect data. Data is collected using the executable code. The collected data values are provided as output. Metrics management is also disclosed. A configuration of a metric is obtained. The metric configuration includes a definition of how computation of the metric is to be performed and a mapping between a computation input and collected data. Collected data values are obtained based at least in part on the mapping. Metric values are computed according to the definition. One or more results associated with the computed metric values are stored.

Abstract: A system for utilizing models derived from offline historical data in online applications is provided. The system includes a processor and a memory storing machine-readable instructions for determining a set of contexts of the usage data, and for each of the contexts within the set of contexts, collecting service data from services supporting the media service and storing that service data in a database. The system performing an offline testing process by fetching service data for a defined context from the database, generating a first set of feature vectors based on the fetched service data, and providing the first set to a machine-learning module. The system performs an online testing process by fetching active service data from the services supporting the media streaming service, generating a second set of feature vectors based on the fetched active service data, and providing the second set to the machine-learning module.

Abstract: Some content players use manifest files to obtain and play video content. The manifest files include metadata information for obtaining video chunks that together constitute a full video, and can be used by content players to obtain the chunks, stitch them together, and render the video as a stream. Described herein are techniques for dynamically generating manifest files. A request for a manifest file is received, such as from a video player. Manifest file contents are retrieved, such as from an original manifest file. A manifest file to be used by a video player to fetch video chunks is then dynamically generated and returned.