Amazon Technologies, Inc.

Ankit Kumar

Alazel Acheson

Gaurav Gupta

Jasmeet Chhabra

Mark Aran Aiken

Matthew William Berry

Raghunathan Kothandaraman

Dan Popick

Sasanka Rajaram

Deep Dixit

Ankul Rastogi

Amit Sahasrabudhe

Dhaval Parmar

Sandesh Sanjeev

Meng Li

Daniel Stephen Popick

Rachit Jain

Marius Popa

Munjal Doshi

Luke Edward Kennedy

David M Hulme

Weixun Wang

Lisa A Winn

Dennis Ono Tjandra

Jamie Layne White

Sharad Sridhar

Iulian Catalin Stafie

The updating of a definition layer or schema for a large distributed database can be accomplished using a plurality of data store tiers. A distributed database can be made up of many individual data stores, and these data stores can be allocated across a set of tiers based on business logic or other allocation criteria. The update can be applied sequentially to the individual tiers, such that only data stores for a single tier are being updated at any given time. This can help to minimize downtime for the database as a whole, and can help to minimize problems that may result from an unsuccessful update. Such an approach can also allow for simplified error detection and rollback, as well as providing control over a rate at which the update is applied to the various data stores of the distributed database.

Multi-tier definition management for distributed data stores

Cross-service usage is dynamically enforced for different network-based services. A cross-service control system may accept requests to specify a usage quota for a client for an interface of a network-based service. A rule to enforce the usage quota may be distributed to request handling nodes for the service. Usage measurements may be collected from the request handling nodes for requests received from the client via the interface. An update to the usage quota may be determined from an evaluation of the usage measurements. The rule at the request handling nodes may be updated to enforce the updated usage quota.

Dynamic cross-service usage enforcement for usage measured across multiple different network-based services

Service interruptions in a multi-tenancy, network-based storage system can be mitigated by constraining the execution of queries. In various examples, a network-based storage system may receive a request to execute a query against data maintained by the network-based storage system. The network-based storage system may perform a unit of work to execute the query, progressing through some, but not all, of a set of operations that are to be completed for completing execution of the query. Upon completion of the unit of work, query execution may be paused, query state data may be saved, and query results may be generated for consumption by the requesting computing device. In some embodiments, tokens that are usable to resume query execution based on the saved query state data may be sent to customer computing devices for resuming query execution on-demand.

Constrained query execution

Techniques for incremental stream switching in a distributed system with flip events are described. A special-purpose flip event is inserted into an event stream that indicates, to recipients of the stream, whether they are to switch to use of an alternate event stream. The flip event may identify which other event stream to use, how to obtain events of that event stream, when to switch to the other event stream, and/or whether that particular recipient is to switch. The recipients may be selectively or iteratively migrated using flip events targeting different recipients, limiting the blast radius of a switch. Via use of a tracking data structure, event replay and event skips can be avoided.

Incremental stream switching in distributed systems

Features are disclosed for managing multiple hosts that stream sequential data to nodes for propagation. The hosts can include a primary host and a failover host that each store a replica of the sequential data. A computing device can monitor the hosts and identify the occurrence of a failover event for the primary host. For example, the computing device may determine the schema has been modified. The computing device can further determine to failover to the failover host based on the failover event. The computing device can insert a flip task into the stream of sequential data for the primary host. The insertion of the flip task may be automatic. Further, each node reading from the stream of sequential data for the primary host may encounter the flip task and failover to the failover host at the same logical position but a different absolute position.

Managing failover between data streams

Techniques for region management for distributed services in provider networks are described. A region management service (RMS) receives requests to enable or disable regions a provider network for users. The RMS identifies what services operate within the involved regions and transmits requests to configure resources within the regions that are specific to the users to the services. The RMS also receives messages indicating whether the transmitting services have successfully configured user resources pertaining to the services for the region.

Region management for distirbuted services in provider networks

Disclosed are various embodiments for filters and event schema for categorizing and processing streaming event data. An event may be generated by a service that describes user interaction a client application executable on a client device. The event may be received as a data structure in a first format. A filter, such as a regular expression filter, may be applied to the data structure to identify an event type for the event. The data structure may be converted from the first format to a second format in accordance with a common event schema able to be interpreted by virtual compute engines.

Filters and event schema for categorizing and processing streaming event data

Disclosed are various embodiments for event processing architecture for real-time user engagement. Events that describe user interactions with client applications are analyzed in real-time when received in a data stream. The events are compared to patterns specified, for example, by administrators. If a pattern is matched based on a comparison to events received in the data stream, a predetermined action may be performed. The processing of events may be performed by a compute engine, which may include a virtual machine or a thread implemented in a parallel computing arrangement.

Event processing architecture for real-time member engagement

Growing community of inventors

Bellevue, WA, United States of America

Ankit Kumar