Amazon Technologies, Inc.

Adam Douglas Morley

Wei Xiao

Swaminathan Sivasubramanian

Michael David Marr

Timothy Andrew Rath

Christopher Nathan Watson

Stefano Stefani

James Christopher Sorenson

Anurag Windlass Gupta

Darin Lee Frink

Samuel James McKelvie

Yan Valerie Leshinsky

David Craig Yanacek

Ian Paul Nowland

Rajaprabhu Thiruchi Loganathan

Steve Huynh

Sharatkumar Nagesh Kuppahally

Kiran Kumar Muniswamy Reddy

Ravi Math

Lokendra Singh Panwar

Owen Cummings

Karl Gohde

Antonio Gabriel Perez De Tejada Martinez

Graham Pedersen

Tate Andrew Certain

Kiran-Kumar Muniswamy-Reddy

David Edward Bryan

Jai Vasanth

Maximiliano Maccanti

Yijun Lu

Rande Anthony Blackman

Vaibhav Jain

Pejus Manoj Das

Xianglong Huang

James Christopher Sorenson, Iii

Barry Bailey Hunter, Jr

Ming-Chuan Wu

Alexander Richard Keyes

Jiandan Zheng

Charles Edward Rice

Glenn A Dierkes

Shirley Xue Yi Ly

Anshul Gupta

Krishnan Seshadrinathan

Vincent Anthony Brancato

Michael D Leonhard

Samuel James Mckelvie

Ian P Nowland

A multi-tenant database may maintain a plurality of datasets on a memory device that is subject to degraded operation caused by a subset of possible state transitions within the device's memory cells. A storage engine may identify entropy characteristics of datasets, independently of other datasets hosted on the memory, and use the entropy to construct a symbol table that maps from data within the dataset to symbols that may be stored on the memory device with a minimized number of state transitions.

Data shaping to reduce memory wear in a multi-tenant database

Consistent queries to replicated data may be performed according to global clock values. Time values for updates performed at a data set may be identified according to a global clock. The updates may be replicated to a replicated portion of a data set. Queries directed to the replicated portion of the data set may be performed in response to determining that the time values for updates indicate that the replicated portion of the data set is consistent with a time value identified for the query according to the global clock.

Global clock values for consistent queries to replicated data

Scalable architecture for propagating updates may be implemented for data replicated from a data set. A node may receive updates to items in a data set that have been committed to the data set. The node may determine whether the update should be applied to a replicated portion of the data set. For updates that should be applied, the node may identify another node that hosts the replicated portion of the data set and send a request to the other node to perform a conditional atomic operation to apply the update to the item in the replicated portion of the data set. The condition may compare a version identifier associated with an update and a current version identifier for the item at the other node. If the condition evaluates true, then the update to the item in the replicated portion may be performed.

Scalable architecture for propagating updates to replicated data

A database system may implement database system recovery using non-volatile system memory. An update to a data page of a database may be received. A version of the data page may be obtained in system memory. A new version of the data page may be determined according to the received update. The new version of the data page may be maintained in a non-volatile portion of system memory irrespective of a failure of the database. In at least some embodiments, the update may be performed without generating recovery log records indicating the update. Upon recovery from a database failure, data pages maintained in non-volatile system memory may be made available to service access requests.

Database system recovery using non-volatile system memory

A system capable of identifying and locking resources (e.g., items, tickets, etc.) may map resources to a bit vector and use bit vectors to efficiently identify and lock the resources. For example, tickets to a concert may be mapped to a bit vector and the system may generate an availability bit vector indicating an availability of the seat locations. Using bit vectors, the system may easily identify available seat locations that satisfy a search request and may select from the available seat locations using an optimization technique. For example, the system may select groups of seats that are next to unavailable seat locations or avoid selling groups of seats that would leave only a single seat available. Thus, the system may avoid fragmentation of ticket sales while improving a latency associated with searching for and reserving seat locations.

Method for selling and searching for items

Methods and apparatus for dedicated endpoint instances for network-accessible services are disclosed. The number of dedicated endpoint instances to be established to direct work requests of a given client to back-end nodes of a service may be determined. Configuration policies for the instances, including one or more policies related to caching, authentication, redundancy, durability, availability, and/or load balancing, may also be determined. Configuration of the instances may be initiated, and an indication that the endpoint or endpoints have been configured may be provided to the client.

Dedicated endpoints for network-accessible services

A system capable of identifying and locking resources (e.g., items, tickets, etc.) may map resources to a bit vector and use bit vectors to efficiently identify and lock the resources. For example, tickets to a concert may be mapped to a bit vector and the system may generate an availability bit vector indicating an availability of the seat locations. Thus, the tickets may be reserved by changing a binary value for bits corresponding to individual tickets. In addition, the system may generate bit vectors corresponding to metadata associated with the seat locations (e.g., aisle seat, handicap accessible, first ten rows, back ten rows, etc.), allowing the system to easily identify available seat locations that satisfy the search request. As the bit vectors are relatively small and can be processed easily, the system may improve a latency and reduce an error rate associated with searching for and reserving seat locations.

A system may maintain a plurality of datasets within a fleet of memories subject to degraded operations caused by state transitions. A dataset maintained on one of the memories may be relocated to another memory. The dataset may be selected based on recorded observations that may be converted into a metric indicative of state transitions associated with maintaining a dataset. A dataset may be selected for relocation based on being associated with a high number of state transitions. The dataset may be relocated to a memory currently having a long operational lifetime relative to other memories in the fleet.

Mechanism for distributing memory wear in a multi-tenant database

A distributed database system may comprise a master computing node and one or more replication peer nodes. Upon identifying a failure of one of the nodes, a control plane may obtain a computing node instance and configure the computing node instance to act as a log-only peer to the remaining nodes. The log-only peer may store the log tail of the master computing node and may participate in quorum decisions. Upon restoration of the failed node, the log-only peer may be returned to the pool of computing node instances.

Dynamic replication peering

Methods and apparatus for resource silos at network-accessible services are disclosed. A subset of resources used for a database service, including at least one resource from each of a plurality of data centers, is selected for membership in a resource silo based on grouping criteria. A silo routing layer node identifies the resource silo as the target silo to which a client work request is to be directed. The client work request is sent to a front-end resource of the target silo either by the client, or by the silo routing layer node on behalf of the client. The front-end resource of the target silo transmits a representation of the work request to a back-end resource of the target silo, where a work operation corresponding to request is performed.

Resource silos at network-accessible services

Application program data stored in system memory may be selectively persisted. An indication may be provided to an application program that an application data object or a range of application data stored in system memory may be treated as persistent. Data backup may be enabled for the application data object or range of application data in the event of a system failure, copying the application data object or range of application data from system memory to non-volatile data storage. Upon recovery from a system failure, further data backup for the application data object or the range of application data may be disabled. In some embodiments, at least some of the application data object or range of application data may be recovered for the application program to access. Data backup for the application data object or the range of application data may also be re-enabled.

Selectively persisting application program data from system memory to  non-volatile data storage

A non-relational data store may implement validating and non-validating secondary indexes for a table. Operations at a table for a given item may be performed when indexing the item to create a secondary index or when updates to the given item are received. Attribute values of a given item may be validated with respect to an indexing schema for the secondary index. For a non-validating secondary index, validation errors detected for the attribute values may be ignored so that the operation at the table may be performed. For a validating secondary index, validation errors detected for the attribute values may result in denying performance of the operation. In some embodiments, a secondary index from may be changed from validating to non-validating, or non-validating to validating.

Validating and non-validating secondary indexes for a table in a non-relational data store

A data storage system may implement implicit prioritization to rate-limit secondary index creation for an online table. A secondary index may be generated for a table stored in a data store. The table may be incrementally indexed, performing multiple indexing operations to populate the secondary index. Prior to performing an indexing operation, an evaluation of a capacity limitation for performing indexing operations may be made with respect to capacity to process access requests at the data store. If a determination is made that performance of the indexing operation exceeds the capacity limitation, then the indexing operation may be throttled. If a determination is made that performance of the indexing operation does not exceed the capacity limitation, then the indexing operation may be performed.

Implicit prioritization to rate-limit secondary index creation for an online table

Growing community of inventors

Seattle, WA, United States of America

Adam Douglas Morley