Amazon Technologies, Inc.

Haifeng He

Akshat Vig

Almero Gouws

Derek William Richard Chen-Becker

Lewis Bruck

Aishwarya Kamal

Stefano Stefani

Wei Xiao

Matthew Shawn Wilson

Sudipta Sengupta

Somasundaram Perianayagam

Poorna Chand Srinivas Perumalla

Pejus Manoj Das

Lakshmi Narasimha Guptha Munuhur Rajagopal

Jiacheng Guo

Benny Tran

Vladimir Mitrovic

Yongcong Luo

Varsha Ganesh

Andrey Markin

Shirley Xue Yi Leung

A distributed database management system comprises a plurality of peers and maintains data on behalf of a plurality of clients. Upon receiving a request to access data, the distributed database management system identifies, from among the plurality of peers, a set of peers associated with the client issuing the request. The distributed database management system provides, to the client, a list of the set of identified peers. Access to data maintained by the distributed database system is based, at least in part, on connections between the client and the set of identified peers.

Peer-based access to distributed database

A distributed database system maintains data for a logical table by storing, on a plurality of storage nodes, a collection of key-item pairs. The distributed database system receives a query of the logical table, and identifies one or more portions of a key specified by the query. Based on the one or more portions of the key, the distributed database causes at least one of a get, range query, or scan operation to be performed by one or more of the storage nodes. Results for the query are generated based on one or more items obtained by performance of the operation.

Query language operations using a scalable key-item data store

A service implements techniques for modifying a data table stored in a key-value or non-relational database via a table-oriented schema. In some aspects, a request to modify a table-oriented schema of a table of data may be received. In response to the request, an association between a customer identifier of elements of the table and an internal identifier of elements of the table may be stored that updates a prior association between a customer identifier and an internal identifier. The table-oriented schema of the table may be modified using one or more keys that are generated using the internal identifier.

Alter table implementation details with schema versioning

A data storage service implements encoding techniques for generating compound partition keys, compound clustering keys, or a combination for more flexibility in storing data in a key-value database system. In some aspects a request to store data with multiple partition keys and/or multiple clustering keys may be received. Length information may be added to the partition keys, and the results concatenated together to form a compound partition key used to store the data in the database. Ordering information may be added to the clustering keys, and the results concatenated together to form a compound clustering key, to be used in storing and sorting data in the database. In some cases, the clustering keys may be made a uniform length, such as by padding keys, as necessary, to have a length that is a multiple of an arbitrary number, and then dividing the clustering keys into chunks of that arbitrary number.

Compound partition and clustering keys

Implementations detailed herein include description of a computer-implemented method to migrate a machine learning model from one accelerator portion (such as a portion of a graphical processor unit (GPU)) to a different accelerator portion. In some instances, a state of the first accelerator portion is persisted, the second accelerator portion is configured, the first accelerator portion is then detached from a client application instance, and at least a portion of an inference request is performed using the loaded at least a portion of the machine learning model on the second accelerator portion that had been configured.

Accelerator based inference service

Growing community of inventors

Bellevue, WA, United States of America

Haifeng He