Amazon Technologies, Inc.

James Christopher Sorenson

Yun Lin

Calvin Yue-Ren Kuo

Jonathan I Turow

David Carl Salyers

Douglas Stewart Laurence

Shyam Krishnamoorthy

Allan Henry Vermeulen

Somasundaram Perianayagam

Ankur Khetrapal

Akshat Vig

Mark Edward Rafn

Colin Laird Lazier

John David Cormie

Kestutis Patiejunas

Marc John Brooker

Jason George McHugh

Marco Argenti

Vaibhav Jain

Weili Zhong McClenahan

Alan Conrad Rawcliffe

Alan B Atlas

Eric Michael Wagner

David M Barth

Gunavardhan Kakulapati

Amit Gupta

Ami K Fischman

Sharatkumar Nagesh Kuppahally

Alexander Richard Keyes

Nishanth Alapati

Venkatraghavan Srinivasan

Stefano Stefani

Kyle Michael Roche

William Alexander Stevenson

Gang He

Neha Shetty

Akhilesh Mritunjai

Cheng Pan

Rishabh Jain

Chase Kernan

Fan Zhang

Akshay Suhas Vaidya

Samuel John Young

Andrew Gary Hourselt

Dumanshu Goyal

Nicholas Gordon

Ravi Math

Kyunghwan Choi

Peter Zhivkov

Saurabh Goel

Tate Andrew Certain

Bryan James Donlan

Paul David Franklin

Douglas Brian Terry

Mark Joseph Cavage

Mark Christopher Seigle

Adam Douglas Morley

Christoph Saalfeld

Ivan Brugiolo

Rajesh Shanker Patel

Oren Yossef

Rudresh Amin

Christian L Claiborn

Arijit Choudhury

Richard Michael Sears

Zhong Ren

Tony Petrossian

Nathan Pellegrom Riley

Ambuj Tandon

Yen-Lin Chen

Dmytro Taran

Mark Robert Sieklucki

Steven Collison

Akanksha Fouzdar

Naveen Kumar Aitha

Remi Paucher

Dave Rahmany

Ardis G Maison

Eric Jason Brandwine

Ron Diamant

Pradeep Vincent

Wei Xiao

Andrew Bruce Dickinson

Craig Wesley Howard

David Craig Yanacek

Yosseff Levanoni

Maximiliano Maccanti

Paul Henry Hohensee

Wayne William Duso

James Caleb Kirschner

Rameez Loladia

Matti Juhani Oikarinen

Rajaprabhu Thiruchi Loganathan

Matteo Frigo

Alex Levin

Claire Elizabeth Suver

Steve Huynh

Ashutosh Thakur

Almero Gouws

Alexandra Elizabeth Baoboe Lee

A response initiation time target based at least in part on a service level agreement is assigned to a task requested from a network-accessible service. A deadline parameter of a thread identified to perform sub-tasks of the task at a service logic implementation node is set based on the target. The thread is scheduled for execution using an operating system thread scheduling algorithm which selects threads for execution based on their deadline parameters. After a first sub-task of thread is completed, a response is sent to the requester of the task.

Latency service level agreement based scheduling of operating system threads at cloud services

Contents of client-initiated handshake messages of a security protocol are obtained at a handshake processing offloader configured for an application. The offloader uses a first security artifact (which is inaccessible from a front-end request processor of the application) and the contents of the handshake messages to generate a second security artifact. The second security artifact is transmitted to the front-end request processor, which uses it to perform cryptographic operations for client-server interactions of the application.

Security protocol handshake offloading

Different types of index structures are used for a replica group of a database. A leader node of a replica group performs receives updates to a copy of the database using a first type of index structure. A follower node performs updates received from the leader node as a log of updates to a copy of the database in an external storage system when a size of the received updates exceeds a threshold. The follower node performs requests to read data from the database using the copy in the external storage.

Different types of index structures for storing database data in a replica group

Systems and methods are provided to implement a fast recovery process in a partitioned replicated data store. In some embodiments, the data store is configured to store an object in a plurality of partitions and replicate data in each partition in a group of replica nodes to satisfy a durability model. In response to a replica failure, the data store performs a split operation to create a plurality of new partitions. The partition's data is split into subsets corresponding to the new partitions. The subsets are transmitted, in parallel, from the surviving replica nodes of the partition to new replica nodes in the new partitions. The new partitions then replicate respective subsets of data in their respective replication groups using a chained replication technique. The recovery process allows the data store to return into compliance with the durability model more quickly, by parallelizing the copying of data.

Fast recovery of partioned replicated data store

A distributed, web-services based storage system. A system may include a web services interface configured to receive, according to a web services protocol, a given client request for access to a given data object, the request including a key value corresponding to the object. The system may also include storage nodes configured to store replicas of the objects, where each replica is accessible via a respective unique locator value, and a keymap instance configured to store a respective keymap entry for each object. For the given object, the respective keymap entry includes the key value and each locator value corresponding to replicas of the object. A coordinator may receive the given client request from the web services interface, responsively access the keymap instance to identify locator values corresponding to the key value and, for a particular locator value, retrieve a corresponding replica from a corresponding storage node.

Distributed storage system with web services client interface

Updates to projected data subsets may be ordered using conditional operations. When updates to a data set are received, a determination is made according to a schema for a projected subset of the data set as to whether the update is to be propagated to the projected subset of the data set. If the update is to be propagated, a comparison of a version identifier for the update may be made with a version identifier for the subset to determine whether performance of the update is consistent with an ordering of updates performed to the first data set. If the comparison indicates the performance of the update is consistent, then the update may be performed to the projected subset. If not, then the update may not be performed.

Ordering updates to secondary indexes using conditional operations

A storage gateway serves as an interface between processes on a customer network and a service provider. The storage gateway is located on-premise with the customer processes. To customer processes, it appears that data is stored locally. However, the storage gateway interfaces with a remote storage service to store the data. For cached gateways, the primary data store is a remote data store, while frequently accessed data may be locally cached by the gateway. Reads may be satisfied from the local cache or from virtual data storage; writes are handled so as to appropriately update data blocks in the local cache and/or in virtual data storage. For shadowing gateways, the primary data store is the local data store; reads are passed through to a local data store, and writes are shadowed to virtual data storage as well as being sent to local data store.

Remote storage gateway management using gateway-initiated connections

Offline building of a projected data subset may be performed. A request to create a data set that is a projected subset of data from a source data set may be received. A data store separate from the data store storing the source data set may store a copy of the source data set that is used to replicate items to the projected subset of data according to a schema for the projected data subset. Updates made to the source data set may also be replicated to the projected data subset according to the schema. Conflicts between replicated items and replicated updates to the projected data set may be resolved by comparing a version identifier for the replicated update and replicated item to determine what to store in the projected data subset.

Offline index builds for database tables

A distributed load balancer in which a router receives packets from at least one client and routes packet flows to multiple load balancer (LB) nodes according to a per-flow hashed multipath routing technique. For a given packet flow, the LB nodes randomly select a server node as a target for the packet flow from among multiple server nodes and send a connection request to the server node. A load balancer module on the server node makes the decision on whether to accept or reject the connection based on one or more metrics indicating a respective server's current load. If the module accepts the connection request, a connection is established between the server and the client. Otherwise, the load balancer nodes may select another server node and try again. The connections established between clients and servers pass through the load balancer nodes, but are not terminated at the load balancer nodes.

Distributed load balancer

A database system may associate functions with a database table. A request to associate a function with a table in a database system may be received. An association between the table and the function may be created. The function may include parameters that are determined from values within the table which are then invoked by a request to perform the function. The associated function may cause the collection of the values prior to performance of the function.

Associating a function with a table in a database system

Replicas may be added to a multi-leader replica group for a data set. A new replica may be added to a group of replicas by copying a version of the data set according to a point in time to the new replica. As updates may occur at different ones of the replicas when the new replica is added, updates that occurred at a source replica for the snapshot and other replicas may be replicated by the different replicas according to the point in time at which the snapshot is created without blocking updates at any of the replicas in the replica group.

Adding replicas to a multi-leader replica group for a data set

A computing environment is disclosed that receives from devices requests directed toward services accessible in the environment, and that forwards communications from services in the environment to devices registered with the environment. During a registration process at the environment, devices are assigned a device identifier that is used to identify and authenticate each particular device and requests communicated from and to the device via the environment. The computing environment maintains state information for each device that has been registered with the system. As the device interacts with the system, the state information is updated to reflect the changes in the device. When requests to perform functions are received from devices, the computing environment determines for the particular device and the particular function requested what processing needs to be performed by the environment in response to the request.

Device state management

A first run-time environment executing a first instance of an application, and a second run-time environment executing a second instance of the application are established. An indication of an impending commencement of a reduced-capacity phase of operation of the first run-time environment is obtained at a service request receiver. Based at least in part on the indication, the service request receiver directs a service request to the second instance of the application.

Service performance enhancement using advance notifications of reduced-capacity phases of operations

Resilient consistent hashing techniques are implemented to perform requests among nodes of a distributed system. Different load balancers are assigned different groups of request handling nodes that cache data to perform request. Requests are assigned to different load balancers according to a consistent hashing scheme. Load balancers distribute requests among assigned nodes using a load balancing scheme. Different numbers of nodes can be assigned to different load balancers.

Resilient consistent hashing for a distributed cache

Methods, systems, and computer-readable media for a scalable metadata index for a time-series database are disclosed. A time-series database receives a query comprising spatial and temporal boundaries of requested time-series data in a table. Using traversal of a metadata index, the time-series database determines one or more storage locations in the one or more data sources corresponding to the spatial and temporal boundaries. The metadata index includes a root node representing spatial and temporal boundaries of the table. The metadata index includes one or more leaf nodes representing the spatial and temporal boundaries of the requested time-series data. The one or more leaf nodes indicate the one or more storage locations in the one or more data sources. The time-series database obtains one or more elements of the requested time-series data from the one or more storage locations in the one or more data sources.

Scalable metadata index for a time-series database

Growing community of inventors

Seattle, WA, United States of America

James Christopher Sorenson