International Business Machines Corporation

Scott Thomas Marcotte

Timothy Peter Krein

Gregory David Doolittle

Kershaw S Mehta

James M Showalter

Geoffrey E Miller

Mary Ng

Susan M Kimmel

A memory-only snapshot of a file is disclosed that can be used by a system to perform a read of the file without disrupting other systems from reading or writing to the file. The in-memory snapshot structure includes a copy of allocation information of the file that points to the file data and be can be used by a backup application to read a file. A lock manager is provided that is configured to include an output parameter for file locks that includes allocation information associated with the in-memory file snapshot, thereby notifying a writer to the file that it must perform copy-on operations. A free space manager is also provided that is configured to track pinned blocks so as to both allow a write to a file to free snapshot blocks, while also preventing the freed blocks from being allocated to other files until backup read processing is completed.

Concurrent writing to a file during backup of the file

Responsive to determining an in-memory image of a metadata disk block, a buffer is generated in memory and includes metadata updates made to the metadata disk block. Metadata updates to the disk block by a first transaction are recorded and stored in the buffer. Transfer of metadata updates that are logically complete, but remain in an active transaction list stored in the buffer, are delayed to a completed transaction list, scheduled to be written to a log file on disk at a subsequent time. A determination is made as to whether the metadata disk block is deleted by a second transaction following the first, and responsive to determining the metadata disk block is to be deleted and the metadata updates of the second transaction transfer to the completed transaction list, preventing the writing of the metadata updates to the log file and the metadata disk block to disk.

Granular buffering of metadata changes for journaling file systems

A cached version of a file system directory is synchronized with a server version of the directory in a distributed file system. Both a server and a client specify version numbers for their directory versions. On receiving a request from the client specifying a directory update, the server updates its version, increments its version number, and transmits a reply to the client containing a change log with the incremented version number. Upon receiving the reply, the client compares the received version number with that of its cached version. If the version number matches that of the next expected update, the client applies the updates to its cached version and increments its version number. Otherwise, it adds the received change log to a change log queue for the directory without incrementing the last applied version. Mechanisms are provided for handling parallel read and update requests without awaiting replies from the server.

Method, apparatus and computer program product for maintaining file system client directory caches with parallel directory writes

A method comprising, sending an instruction from a validator member of a sysplex to a second member of a sysplex to prevent operations that alter a sysplex namespace, determining whether the validator member of the sysplex is communicative with the second member of the sysplex and whether the members of the sysplex retain current sysplex status information, sending an instruction from the validator member to the second member to send a first namespace attribute data associated with the second member to the validator member responsive to determining that the validator member of the sysplex is communicative with the second member of the sysplex and the members of the sysplex retain current sysplex status information, determining whether the first namespace attribute data associated with the second member is consistent with a second namespace attribute data, and sending an instruction to correct the first namespace attribute data.

Validation and correction in a distributed namespace

Deadlock situations within a computing environment are avoided by properly managing pools of threads used to service requests of the computing environment. When a server of the computing environment receives a request to be processed and that request is waiting on a response from a client of the computing environment, the set of eligible thread pools for the response is dynamically altered. This dynamic altering allows the response to be serviced by a thread pool different from the thread pool servicing the request, thereby avoiding a deadlock situation.

Method, system and program products for managing thread pools of a computing environment to avoid deadlock situations

Data movement within a computing environment is at the very least reduced. Data is transmitted between a file system of the computing environment and a transmission medium of that environment. The transmission includes bypassing non-file system buffers in performing the transmission. For example, when data is sent to the file system to be written to one or more storage media, the file system swaps one or more buffers of the file system with the one or more buffers containing the data. The swapping does not require the copying of data. Further, for a read operation, the file system calls a routine, which is provided with one or more pointers to the data that is to be sent to a requester of the data.

Method, system and program products for reducing data movement within a computing environment

Method, system and program products for reducing data movement within a computing environment by bypassing copying data between file system and non-file system buffers in a server

Method, system and program products for managing thread pools of a computing environment to avoid deadlock situations by dynamically altering eligible thread pools

Throughput of a request (e.g., a write request) is enhanced by allowing a receiver of the request to reply to the request before fully receiving the entire request. This reduces the delay between the time that the receiver completes a request and the time that the sender sees that the request is finished, and allows other requests to be forwarded prior to completion of the first request. The other requests that are forwarded, however, are processed in the order of receipt. Additionally, any errors that occur in a request after the reply is sent is forwarded to the sender of the request, even though the reply has previously been sent.

Method, system and program product for managing the flow of data between senders and receivers of a computing environment

Tasks make updates requested by calling tasks to a shared resource serially in a first come first served manner, atomically, but not necessarily synchronously, such that a current task holding an exclusive lock on the shared resource makes the updates on behalf of one or more calling tasks queued on the lock. Updates waiting in a queue on the lock to the shared resource may be made while the lock is held, and others deferred for post processing after the lock is released. Some update requests may also, at the calling application's option, be executed synchronously. Provision is made for nested asynchronous locking. Data structures (wait_elements) describing update requests may queued in a wait queue for update requests awaiting execution by a current task, other than the calling task, currently holding an exclusive lock on the shared resource. Other queues are provided for queuing data structures removed from the wait queue but not yet processed; data structures for requests to unlock or downgrade a lock; data structures for requests which have been processed and need to be returned to free storage; and data structures for requests that need to be awakened or that describe post processing routines that are to be run while the lock is not held.

Interface system and method for asynchronously updating a share resource with locking facility

In a computing environment having clients with different semantics or protocols, a capability is provided that enables those clients to share the same data or files. A token management function is provided that allows clients that did not previously support token management to use the token management function to access the shared files. These capabilities are provided without requiring modifications to the client software.

Method, system and program products for sharing data between varied clients using token management

System and method for scheduling I/O requests in a multi-tasking data processing environment. An I/O request issued by an application is placed in an I/O request holding queue. Under control of the requesting application (or, alternatively, the operating system), the I/O request is selectively canceled or moved to a service pending queue for execution. Requests can be moved either by the application or by the Operating system when an I/O completes (and hence the service pending queue has room for another IO).

Growing community of inventors

Vestal, NY, United States of America

Scott Thomas Marcotte