Emc IP Holding Company LLC

Hewlett-packard Development Company, L.p.

Hewlett Packard Enterprise Development LP

Maginatics LLC

Niraj Tolia

Vaibhav Kamra

Nathan Rosenblum

Diwaker Gupta

Parthasarathy Ranganathan

Rajiv Desai

Deepti Chheda

Andrew Gaul

Cullen Edwin Bash

Manish Marwah

Jichuan Chang

Yoshio Frank Turner

Nathan Lorenzo Binkert

Zhikui Wang

Julio Lopez

Thomas E Sandholm

Mahadev Satyanarayanan

Vanish Talwar

Xiaoyun Zhu

Jose Renato G Santos

Shrinand Javadekar

Vijay Panghal

Partha Ranganathan

Shailesh N Joshi

Andrew James Phelan

Biju Shanmugham Pillai

Gunho Lee

Vishakha Gupta

Luca Parolini

Bruno Sinopoli

Bruce Krogh

Compounding file system metadata operations via buffering is disclosed. In various embodiments, an indication to send locally-stored file system operations to a remote file system metadata server is received. A dependency graph or other data structure reflecting dependencies between the locally-stored file system operations to identify a linear chain of dependent file system operations. Operations comprising the linear chain are combined into a compound file system operation. The compound file system operation is sent to the remote file system metadata server.

Compounding file system metadata operations via buffering

A method, article of manufacture, and apparatus for limited deduplication scope on a distributed file system is discussed. A write request is received from a client at the metadata server ('MDS'), where the write request comprises a data object identifier and a preferred object store identifier. The MDS determines whether a preferred object store associated with the preferred object store identifier contains a copy of a data object associated with the data object identifier. A write URL comprising the data object identifier and a object store location associated with the preferred object store is transmitted to the client when the preferred object store does not contain the copy of the data object.

Limited deduplication scope for distributed file systems

Cancelling file system operations at a file system client of a distributed file system is disclosed. In various embodiments, an indication is received to send locally-stored file system operations to a remote file system metadata server. A dependency graph or other data structure reflecting dependencies between the locally-stored file system operations is used to identify a set of locally-stored file system operations the members of which are fully dominated by a create-delete pair of operations comprising the set. The locally-stored file system operations comprising the set are cancelled, without being sent to the remote file system metadata server.

File system operation cancellation

Techniques to eliminate redundant file system operations are disclosed. In various embodiments, an indication is received to send locally-stored file system operations to a remote file system metadata server. A set of related locally-stored file system operations is determined. Operations in the set are determined to satisfy a merger criteria and are merged into a single merged file system operation. The merged file system operation is sent to the remote file system metadata server.

Eliminating redundant file system operations

Techniques to manage the use of hash functions are disclosed. In various embodiments, a hash function epoch metadata is stored to associate a hash function with an epoch, the epoch defining a subset of data stored by a distributed file system. The distributed file system is configured to use the hash function epoch metadata to manage storage of data comprising the subset of data stored by the distributed file system.

Handling weakening of hash functions by using epochs

Providing disaster recovery of a distribute file system metadata server as a service is disclosed. In various embodiments, an indication is received to provide alternative access to a file system metadata server. A copy of a self-describing backup of the file system metadata server is obtained, e.g., from cloud-based storage. The backup is used to provide access to a read only instance of the file system metadata server.

Disaster recovery as a service

Systems and methods for protecting and deduplicating streams of data in a cloud based platform. A platform is configured with multiple services and can communicate with multiple clients. The platform receives all requests at an endpoint and distributes the requests to workers using multiple queues. The platform may be stateless and is scalable. The meta-data is handled separately from the data in one example.

Scalable cloud hosted metadata service

A method, article of manufacture, and apparatus for providing a site cache in a distributed file system is discussed. Data objects may be written to a site cache prior to writing them to an authoritative object store. This provides performance benefits when a client writing the data has a better connection to the site cache than to the authoritative object store.

Site cache for a distributed file system

Illustrated is a system and method that includes identifying a search space based upon available resources, the search space to be used to satisfy a resource request. The system and method also includes selecting from the search space an initial candidate set, each candidate of the candidate set representing a potential resource allocation to satisfy the resource request. The system and method further includes assigning a fitness score, based upon a predicted performance, to each member of the candidate set. The system and method also includes transforming the candidate set into a fittest candidate set, the fittest candidate set having a best predicted performance to satisfy the resource request.

Computing cluster performance simulation using a genetic algorithm solution

Managing non-congestion-controlled message traffic in a datacenter. Any non-congestion-controlled packets originating in a sending server and intended for a receiving server are encapsulated in the sending server and transmitted through a virtual circuit having a message transport protocol with rate control. The packets are decapsulated in the receiving server.

Methods of and apparatus for managing non-congestion-controlled message traffic in a datacenter

A method in data storage involves receiving at a metadata server from a client device a BEGIN_WRITE request identifying data to be written by hashes, consulting, by the metadata server a deduplication table to determine historical nature of the data determined to be written, determining that the data was never previously written, was previously written by another client, or was previously written but failed a verification, sending to the client device write locations with tokens for the data, starting an authentication timer, and authorizing the client device to write directly to the data server using the authentication tokens and the authenticated write locations, determining that the data was previously written by the same client, or that the data was previously written and passed the verification test, notifying the client of this circumstance, and receiving from the client device an END_WRITE request.

Out-of-path, content-addressed writes with untrusted clients

A system and method is shown that includes an admission control module that resides in a management/driver domain, the admission control module to admit a domain that is part of a plurality of domains, into the computer system based upon one of a plurality of accelerators satisfying a resource request of the domain. The system and method also includes a load balancer module, which resides in the management/driver domain, the load balancer to balance at least one load from the plurality of domains across the plurality of accelerators. Further, the system and method also includes a scheduler module that resides in the management/driver domain, the scheduler to multiplex multiple requests from the plurality of domains to one of the plurality of accelerators.

Growing community of inventors

Sunnyvale, CA, United States of America

Niraj Tolia