Hewlett-packard Development Company, L.p.

Hewlett Packard Enterprise Development LP

Amazon Technologies, Inc.

Joseph Andrew Tucek

Eric A Anderson

Anurag Windlass Gupta

Harumi Kuno

Mark David Lillibridge

Saileshwar Krishnamurthy

Alexandre Olegovich Verbitski

Kamal Kant Gupta

Tengiz Kharatishvili

Xiaozhou Li

John Johnson Wylie

Bohan Liu

Goetz Graefe

Mustafa Uysal

Yi Ding

Charles Stuart Johnson

Murali Brahmadesam

Sandor Loren Maurice

Punit Rajgaria

Saleem Mohideen

Changan Han

Xiaofeng Bao

Alexey Kuznetsov

Alok Nandan Nikhil

Niket Goel

Milind M Chabbi

Lawrence Douglas Webley

Debanjan Saha

Mushahid Alam

Tiffany Jianto

Mesut Kuscu

Onkar Patil

Tuan Tran

William Scherer

Jain Nitin

Optimistic concurrency may be implemented for multiple writers to a database. Writes to a database from different writers may succeed unless a conflict is detected at the data store for the database. For detected conflicts, a request to resolve the conflict between writes may be submitted to a primary writer, which may determine and provide a resolution of the conflict to the conflicting writers. The data store may then be updated to commit the selected write identified by the resolution of the conflict.

Optimistic concurrency for a multi-writer database

A service provider network implements seamless scaling via proxy replay of session state. Upon a trigger, such as a determination to scale a server, a scaled server may be spun up and an identifier of the scaled server provided to a first (existing) server. The first server sends the identification of the second server, and session state information for each of the connections between the first server and the request router, to the request router. For each of the connections, the request router establishes a new connection between the request router and the second (scaled) server, and replays the session state information for the connection to the second server. The request router then routes traffic between each existing client connection (e.g., the same existing client connection which carried traffic delivered to the first server) and the corresponding new connection to the second server.

Seamless scaling via proxy replay of session state

A distributed system implements initializing reliability data of a newly joined node of a cluster for purposes of leadership election. When a new node joins a cluster, the new node receives reliability data for one or more existing nodes of a cluster. The new node initializes its own reliability data based on the reliability data received from the existing nodes of the cluster. In some cases, the new node may calculate a median reliability of the existing nodes and then initialize its own reliability to be the median or slightly below the median. For example, if the median reliability is 'two failures,' then the new node may set its reliability to be 'three failures.' This may provide a more stable leadership election process, as the new nodes are unlikely to be elected as a leader node.

Initializing node reliability for leadership election

A high performance computing system including processing circuitry and a shared fabric memory is disclosed. The processing circuitry includes processors coupled to local storages. The shared fabric memory includes memory devices and is coupled to the processing circuitry. The shared fabric memory executes a first sweep of a stencil code by sequentially retrieving data stripes. Further, for each retrieved data stripe, a set of values of the retrieved data stripe are updated substantially simultaneously. For each retrieved data stripe, the updated set of values are stored in a free memory gap adjacent to the retrieved data stripe. For each retrieved data stripe, the free memory gap is advanced to an adjacent memory location. A sweep status indicator is incremented from the first sweep to a second sweep.

Sequential memory access on a high performance computing system

Systems and methods associated with latch-free searching are disclosed. One example method includes receiving a key identifying data to be retrieved from a tree-based data structure. The method also includes performing a concurrent, latch-free search of the tree-based data structure until a leaf node is reached. The method also includes validating the leaf node. The method also includes retreading a portion of the search if the leaf node fails validation.

Latch-free concurrent searching

A computer system for controlling access to a shared data structure includes a shared memory coupled to first and second processing units that stores a multi-lock to control access to a shared data structure. The multi-lock includes a first sub-lock associated with the first processing unit and a second sub-lock associated with the second processing unit The system also includes a data access control engine to receive a request to read from the data structure from the first processing unit and, as a result, determine whether a privately modifiable copy the first sub-lock exists in a first cache dedicated to the first processing unit, acquire a read portion of the first sub-lock and not communicate the acquisition across a coherence bus if a privately modifiable copy of the first sub-lock exists in the first cache, and if a privately modifiable copy of the first sub-lock does not exist in the first cache, load the first sub-lock into the first cache if no copy is in the first cache, shootdown other copies of the first sub-lock, and acquire the read portion of the first sub-lock.

Systems and methods for controlling access to a shared data structure with reader-writer locks using multiple sub-locks

Systems and methods of reducing bandwidth usage of a mobile client are disclosed. An example method may include caching a first version of network content in a mobile client. The method may also include comparing the cached content with a second version of the network content. The method may also include generating a recipe to construct the second version of the network content from the cached network content based on a result of the comparing. The method may also include sending the recipe to the mobile client.

Reducing bandwidth usage of a mobile client

A method for dynamically partitioning a B-tree data structure, includes: determining if the B-tree data structure requires a partition; establishing a midpoint of the B-tree data structure; migrating from a beginning of the B-tree data structure to the midpoint of the B-tree data structure to a first B-tree data structure; migrating from the midpoint of the B-tree data structure to an end of the B-tree data structure to a second B-tree data structure; and allowing normal operations on the B-tree data structure during migration.

Dynamic partitioning of a data structure

A method for determining priority of recovery for a RAID implementation includes detecting a first failure of the RAID implementation; detecting a second failure of the RAID implementation; assigning a first priority to the first failure and a second priority to the second failure; and setting the priority of the recovery based on the first priority and the second priority.

Prioritizing recovery in a storage system implementing raid

A method for inserting a file in a search tree (B-tree) implemented on a file system, includes: (a) in response to a current node being a root node or an internal node, determining a child node; (b) repeating (a) until a leaf node is detected; (c) in response to a number of leaf nodes at a level of the leaf node exceeding an upper limit of files or a number of child nodes of a current node exceeding an upper limit of sub-directories, balancing the level of the leaf node or child nodes; and (d) inserting the file at the level of the leaf node.

File system management and balancing

A system includes storage media and control logic coupled to the storage media, where the control logic is configured to receive a write request and determine whether the write request specifies writing a predetermined pattern to a particular location of the storage media. In response to determining that the write request specifies writing the predetermined pattern to the particular location, the control logic is configured to identify with an indicator that the particular location contains invalid data.

Identifying a location containing invalid data in a storage media

A method and a corresponding apparatus provide for remapping for wear leveling of a memory. The method is implemented as logic and includes the steps of receiving a memory operation, the memory operation including a logical memory address; dividing the logical address into a logical block address portion, a logical line address portion, and a logical subline address portion; translating the logical block address portion into a physical block address; selecting a line remap key; applying the line remap key to the logical line address portion to produce a physical line address; producing a physical subline address portion; and combining the physical block, line, and subline address portions to produce a physical address for the memory operation.

Remapping for memory wear leveling

A method is provided for efficiently recovering information in a distributed storage system where a list of values that should be stored on a storage device is maintained. A first convergence round is scheduled to be performed on the list of values to bring each value to an At Maximum Redundancy (AMR) state. A second convergence round is scheduled to be performed on the list by selecting a wait time interval from a predefined range of wait time intervals between starts of convergence rounds.

Efficient failure recovery in a distributed data storage system

A server receives information that identifies which chunks are stored in local caches at client computers and receives a request to evict a chunk from a local cache of a first one of the client computers. The server determines whether the chunk stored at the local cache of the first one of the client computers is globally oldest among the chunks stored in the local caches at the client computers, and authorizes the first one of the client computers to evict the chunk when the chunk is the globally oldest among the chunks stored in the local caches at the client computers.

Growing community of inventors

Palo Alto, CA, United States of America

Joseph Andrew Tucek