Hewlett-packard Development Company, L.p.

Google Inc.

Hewlett-packard Company

Hewlett Packard Enterprise Development LP

Arif Abdulhusein Merchant

Mustafa Uysal

Kimberly Keeton

Alistair Veitch

Svend Frolund

John Wilkes

Alexander Shraer

Guillermo Alvarez

Brian Frank Cooper

Marcos Kawazoe Aguilera

Atul Adya

Mehul A Shah

Dirk Manfred Beyer

Richard Andrew Golding

Yasushi Saito

Mirjana Spasojevic

Cipriano A Santos

Ludmila Cherkasova

Christos Karamanolis

Ram Swaminathan

Pradeep Padala

Eric A Anderson

James S Jackson

James Michael Reuter

Sharad Singhal

Ajay Gulati

Xiaoyun Zhu

Zhikui Wang

Sumanesh Samanta

Satish Kumar Mopur

Anna Povzner

Alex Xin Zhang

Dinkar Sitaram

Yin Wang

Stavros Harizopoulos

Susan Spence

Murray M Stokely

Yasusuhi Saito

Artyom Sharov

Elizabeth Lynn Borowsky

Alistar Veitch

Kfir Lev-Ari

Vishesh Khemani

Ralph Attila Becker-Szendy

Jeffrey Scott Chase

Mahesh Kallahalla

Lluis Pamies-Juarez

Craig A Soules

Christopher Lumb

Pat Jacobson

Ayman Abouelwafa

Kishore Kumar Muppirala

Christian Eric Schrock

Nina Mishra

Elizabeth A Shriver

Charles B Morrey, Iii

Xudong Shi

David Trastour

Christoph Albrecht

Ernesto Brau

Kenneth Salem

Qian Wang

Tim Sullivan

Francisco Maturana Sanguineti

Stavros Hartzopoulos

Mohit Saxena

Kiran-Kumar Muniswamy-Reddy

Tzu-Wei Yang

Ralph Becker Szendy

A method for optimizing data storage includes obtaining a data object for storage at memory hardware in communication with data processing hardware. The memory hardware includes a plurality of storage devices, each storage device of the plurality of storage devices including storage parameters different from each other storage device of the plurality of storage devices. The method also includes determining one or more data object parameters associated with the data object and predicting, using a model and the data object parameters and the storage parameters, an object temperature representative of a frequency of access for the data object and an object lifetime representative of an amount of time the data object is to be stored. The method further includes selecting, using the predicted object temperature and object lifetime, one of the storage devices, and storing the data object at the selected one of the storage devices.

Optimizing data placement based on data temperature and lifetime prediction

Replicas are selected in a large distributed network, and the roles for these replicas are identified. In one example, a leader is selected from among candidate computing clusters. To make this selection, an activity monitor predicts or monitors the workload of one or more clients. Different activities of the workload are given corresponding weights. The delay in performing requested activities, modified by these weights is found, and the candidate leader with the lowest weighted delay is selected as the leader.

Distributed database configuration

A method includes receiving trace data representing access information about files stored in a large-scale distributed storage system, identifying file access patterns based on the trace data, receiving metadata information associated with the files stored in the large-scale distributed storage system, and generating a preferred storage parameter for each file based on the received metadata information and the identified file access patterns. The method also includes receiving, file reliability or accessibility information of a new file, determining whether the received file reliability or accessibility information of the new file matches information of a file group of the files in the large-scale distributed storage system, and when the file reliability or accessibility information of the new file matches the information of the file group, storing the new file in the large-scale distributed storage system using the preferred storage parameter associated with the file group.

Programmatically choosing preferred storage parameters for files in large-scale distributed storage systems

Methods, systems, and apparatus for automatic sharding and load balancing in a distributed data processing system. In one aspect, a method includes determining workload distribution for an application across worker computers and in response to determining a load balancing operation is required: selecting a first worker computer having a highest load measure relative to respective load measure of the other work computers; determining one or more move operations for a partition of data assigned to the first worker computer and a weight for each move operation; and selecting the move operation with a highest weight the selected move operation.

Weighted auto-sharing

Replicas are selected in a large distributed network, and the roles for these replicas are identified. In one example, a leader is selected from among candidate computing dusters. To make this selection, an activity monitor predicts or monitors the workload of one or more clients. Different activities of the workload are given corresponding weights. The delay in performing requested activities, modified by these weights is found, and the candidate leader with the lowest weighted delay is selected as the leader.

A predictive method for scheduling of the operations is described. The predictive method utilizes data generated from computing an expected lifetime of the individual files or objects within the container. The expected lifetime of individual files or objects can be generated based on machine learning techniques. Operations such as garbage collection are scheduled at an epoch where computational efficiencies are realized for performing the operation.

Optimizing garbage collection based on survivor lifetime prediction

The technology is directed to cache admission control. One or more processors may categorize access requests for data items in a cache storage into a plurality of categories and collect information on the access requests over time. Based on the collected information, a utility value for caching data items in each category of the plurality of categories may be determined. Newly requested data items may be admitted into the cache storage in an order according to the corresponding utility values of their respective categories.

Optimal admission control for caches

Weighted auto-sharding

A transactional memory system uses a volatile memory as primary storage for transactions. Data is selectively stored in a non-volatile memory to impart durability to the transactional memory system to allow the transactional memory system to be restored to a consistent state in the event of data loss to the volatile memory.

Imparting durability to a transactional memory system

Growing community of inventors

Los Altos, CA, United States of America

Arif Abdulhusein Merchant