Microsoft Technology Licensing, LLC

Daniel Edward Lovinger

David Matthew Kruse

Mathew George

James T Pinkerton

Roopesh C Battepati

Mingdong Shang

Thomas Ewan Jolly

Ramesh Shankar

Paul R Swan

Tom Jolly

Supriya Wickrematillake

Norbert Paul Kusters

Neal Robert Christiansen

Sarosh Cyrus Havewala

Ravisankar V Pudipeddi

Keith Stuart Kaplan

David Golds

Scott Colville

Brian D Andrew

Christian Gregory Allred

Matthew Stuart Garson

Chien-Lung Yu

Patrick E Bozeman

A Perry Clarke

Neil R Kaethler

L Felipe Cabrera

Mukul Gupta

Thomas E Jolly

Systems and methods are disclosed for clients and servers operating in a scaled cluster environment. Efficiencies are introduced to the process of connecting a client to a clustered environment by providing the client with the ability to attempt a connection with multiple servers in parallel. Servers operating the in the clustered environment are also capable of providing persistent storage of file handles and other state information. Ownership of the state information and persistent handles may be transferred between servers, thereby providing clients with the opportunity to move from one server to another while maintaining access to resources in the clustered environment.

SMB2 scaleout

Described are embodiments directed at persistent handles that are used to retain state across network failures and server failovers. Persistent handles are requested by a client after a session has been established with a file server. The request for the persistent handle includes a handle identifier generated by the client. The server uses the handle identifier to associate with state information. When there is a network failure or a server failover, and a reconnection to the client, the handle identifier is used to identify replayed requests that if replayed would create an inconsistent state on the server. The replayed requests are then appropriately handled.

Transparent failover

Current file systems may implement opportunistic locks on files that allow clients to locally cache file data, while promoting data consistency when multiple clients read/write the same file. Unfortunately, when a client locally caches directory content of a directory, the file system may lack a mechanism for notifying the client of changes to the directory that would render the cached directory content 'stale'. Accordingly, one or more systems and/or techniques for managing a directory oplock are disclosed herein. A directory oplock may be generated based upon an oplock request comprising a first target key and a first parent key. Subsequent access requests for access to the directory or objects therein may be compared with the directory oplock to determine whether to break the directory oplock, and notify the client (e.g., directory oplock is broken if the client submitting the access request is different than the client owning the directory oplock).

Directory oplock

A directory services implementation is provided to associate distributed file system (DFS) links with individual directory objects, and metadata to attributes thereof, to allow leveraging directory services features for DFS for a given namespace. For example, updating directory objects with modified metadata related to DFS links requires only that the directory object related to the link be updated rather than an entire directory object related to the corresponding namespace. Moreover, directory services functionalities such as querying can be utilized with DFS to provide efficient location of updated records. In this regard, efficient replication of DFS throughout a network is facilitated.

Efficient updates for distributed file systems

A system and method for organizing and sorting targets received in a referral response and for realizing a target fail-back and a target priority policy in a distributed file system is provided. In one embodiment, a sorting method includes receiving a referral response in the form of a list of targets that are sorted into bounded sets. Having a sorted referral response in bounded sets provides a basis for implementing a target fail-back and a target priority policy. The computer system may select a target from a sorted list of targets sorted according to site-cost and/or target priority. Then, the computer system may determine if the set target is associated with a more preferred target when compared to all available targets in the sorted list and if not, switch back to a more preferred target.

System and method for determining target failback and target priority for a distributed file system

A snapshot system capable of capturing snapshots of multiple volumes wherein the snapshots are coordinated. A snapshot manager determines which volumes are to be involved in a snapshot operation, and issues a message to the file system for each volume involved, each message including information sufficient to identify the volumes involved in the snapshot operation. Each file system passes its respective message down to a coordinator mechanism. The coordinator mechanism coordinates the state of each of the volumes, such as by holding writes thereto, to put each volume into a quiescent state, and then enables the snapshot to be captured. When the snapshots are captured, a snapshot set will include snapshots that are coordinated across the multiple volumes. The coordinator mechanism releases any writes being held for the volumes involved.

Kernel-based crash-consistency coordinator

A mechanism is provided that allows an application program to write, as a single file, a large block of data comprising multiple portions that could otherwise be written as several smaller files, then to access, as individual files, each of the portions within the large block of data, and to be able to create individual files efficiently out of each of these subfiles. The mechanism may be partially embodied in a file system that includes an information store defining each file on a volume. The application program writes, via the file system, a single file to the volume as (preferably) a contiguous block of data. The single file includes two or more separable streams of data capable of being stored as individual files (subfiles). Once the single file is written to the volume, multiple entries are made to the information store. Each entry defines and points to a subfile within the single file. The subfiles may be positioned within the single file such that the beginning of each subfile lies on the beginning of an allocation unit. In this manner, the single file may be written to the volume in one efficient operation, yet each subfile is individually accessible via its respective entry in the information store.

System and method for creating multiple files from a single source file

Growing community of inventors

Seattle, WA, United States of America

Daniel Edward Lovinger