John Baker

Based on an analysis of the workload for a key part of a client request to a first service, a need to expand resources is identified. Resource managers from several other services, which help the first service respond to requests, start tasks to provide more resources. These tasks happen independently of each other. Once the tasks are done, a limit for the key is updated to a new value. This new limit is then used to decide whether to accept another client request related to the key.

A resource from a dialog-driven management service is given to a first set of requests when the number of capacity indicators in a throttling data structure goes above a certain level. While the resource is allocated for the first set of requests, one or more rounds of capacity indicator deductions happen. During each round, a certain number of capacity indicators are taken out of the throttling data structure based on a throttling setting. A second set of requests is rejected based on the… Show more

A resource from a dialog-driven management service is given to a first set of requests when the number of capacity indicators in a throttling data structure goes above a certain level. While the resource is allocated for the first set of requests, one or more rounds of capacity indicator deductions happen. During each round, a certain number of capacity indicators are taken out of the throttling data structure based on a throttling setting. A second set of requests is rejected based on the number of capacity indicators left in the throttling data structure. Show less

Techniques for speech recognition are described. For example, a speech recognition service gets a request to perform speech recognition on speech data from a chatbot using a specific speech recognition service. The service identifies a group of hosts to send the speech data to, with the group hosting several speech recognition services including the specific one. The service then determines a path to the group of hosts using certain rules and identifies a particular host to perform the speech… Show more

Techniques for speech recognition are described. For example, a speech recognition service gets a request to perform speech recognition on speech data from a chatbot using a specific speech recognition service. The service identifies a group of hosts to send the speech data to, with the group hosting several speech recognition services including the specific one. The service then determines a path to the group of hosts using certain rules and identifies a particular host to perform the speech recognition. This host, which has the speech recognition service in memory, processes the request and is preferred for performing the task. The speech data is then routed to this specific host, which performs the speech recognition and provides a text result of the speech recognition. Show less

This patent relates to disambiguate between different names for a VUI call request based on the form of communication used. For example, when a user speaks a call request (“Alexa, drop in on kitchen”) the system figures out the request corresponds to a specific mode of communication (e.g., Alexa Comms call, regular telephone call, etc.) and finds both contact data and device data that may match the target name (e.g., “kitchen” may refer to a home device or a restaurant name). The system… Show more

This patent relates to disambiguate between different names for a VUI call request based on the form of communication used. For example, when a user speaks a call request (“Alexa, drop in on kitchen”) the system figures out the request corresponds to a specific mode of communication (e.g., Alexa Comms call, regular telephone call, etc.) and finds both contact data and device data that may match the target name (e.g., “kitchen” may refer to a home device or a restaurant name). The system determines confidence values with respect to each option as well as the form of communications to be used. For example, you may be able to drop in on the device named kitchen, but not the restaurant named kitchen. As such, the system uses the form of communications are relevant context to select the device to call. Show less

A system for testing candidate code to determine if the candidate code is approved for deployment to a production environment. The system may include a code testing engine configured to test a first code set of previously approved code to a testing server and a production environment including multiple production servers. At a deployment time, the code testing engine may then deploy candidate code to the testing server, while the production servers execute the first code set. Performance data… Show more

A system for testing candidate code to determine if the candidate code is approved for deployment to a production environment. The system may include a code testing engine configured to test a first code set of previously approved code to a testing server and a production environment including multiple production servers. At a deployment time, the code testing engine may then deploy candidate code to the testing server, while the production servers execute the first code set. Performance data for a time period before the deployment time and after the deployment time is collected. Latency data sets are fetched from the performance data sets and compared using multiple statistical tests. If the multiple statistical tests generate passing results, the candidate code is approved for deployment to the production environment. Show less