Abstract: The disclosure provides for fault tolerant parallel journaling that speeds up both input/output (I/O) operations and recovery operations. Journal entry writing may occur in parallel with data writing operations. Even if a crash occurs during a data writing operation for which the journal entry has been written, the recovery operation will correctly determine that the journal entry is not valid. Additionally, recovery operations may need to validate fewer journal entries, and yet possibly retain more valid data. Examples include: for each of a plurality of journal entries: receiving incoming data; determining a signature for the incoming data; generating the journal entry for the incoming data; writing the signature in the journal entry; and writing the journal entry and the incoming data to a storage media; and based at least on writing data to the storage media, updating an awaiting index in a journal header.

Abstract: The disclosure supports both trickle and burst input/output (I/O) admission rates in journaling file systems. Examples include receiving incoming data; based at least on receiving the incoming data, generating metadata for a journal entry; adding the metadata to an active metadata batch; issuing a data write to write the incoming data to a storage medium; monitoring for a first trigger comprising determining that a data write for an entry in the active metadata batch is complete; based at least on the first trigger, closing the active metadata batch; and issuing a journal write to write entries of the active metadata batch to the storage medium. A second trigger comprises determining that a batch open time exceeds a selected percentage of a moving average of data write durations. A third trigger comprises determining that a batch counter exceeds a count threshold. These triggers work together to reduce I/O latencies.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for reclaiming one or more portions of storage resources in a computer system serving one or more virtual computing instances, where the storage resources in the computer system are organized in clusters of storage blocks. In one aspect, a method includes maintaining a respective block tracking value for each storage block that indicates whether a call to reclaim the storage block is outstanding; determining, from the block tracking values, a respective cluster priority value for each of the clusters based on a count of storage blocks in the respective cluster for which a call to reclaim is outstanding; and reclaiming a first portion of storage resources in the computer system in accordance with the cluster priority values.

Abstract: The disclosure herein describes managing a rate of processing unmap requests for a data storage volume. Unmap requests are received from a cluster of active hosts that are associated with the data storage volume. Latency data values of each active host are then accessed. A long-term cluster latency average value is calculated based on the accessed latency data values of all active hosts over a long-term time period and a short-term cluster latency average value is calculated based on the accessed latency data values of all active hosts over a short-term time period. An unmap rate adjustment value is calculated based on a difference between the long-term cluster latency average value and the short-term cluster latency average value. The rate of processing unmap requests for the data storage volume is adjusted based on the unmap rate adjustment value and the unmap requests are performed based on the adjusted rate.

Abstract: The disclosure supports both trickle and burst input/output (I/O) admission rates in journaling file systems. Examples include receiving incoming data; based at least on receiving the incoming data, generating metadata for a journal entry; adding the metadata to an active metadata batch; issuing a data write to write the incoming data to a storage medium; monitoring for a first trigger comprising determining that a data write for an entry in the active metadata batch is complete; based at least on the first trigger, closing the active metadata batch; and issuing a journal write to write entries of the active metadata batch to the storage medium. A second trigger comprises determining that a batch open time exceeds a selected percentage of a moving average of data write durations. A third trigger comprises determining that a batch counter exceeds a count threshold. These triggers work together to reduce I/O latencies.

Abstract: The disclosure provides for fault tolerant parallel journaling that speeds up both input/output (I/O) operations and recovery operations. Journal entry writing may occur in parallel with data writing operations. Even if a crash occurs during a data writing operation for which the journal entry has been written, the recovery operation will correctly determine that the journal entry is not valid. Additionally, recovery operations may need to validate fewer journal entries, and yet possibly retain more valid data. Examples include: for each of a plurality of journal entries: receiving incoming data; determining a signature for the incoming data; generating the journal entry for the incoming data; writing the signature in the journal entry; and writing the journal entry and the incoming data to a storage media; and based at least on writing data to the storage media, updating an awaiting index in a journal header.

Abstract: The systems described herein are configured to enhance the efficiency of memory usage and access in a VM file system data store with respect to allocating memory in large and small file block clusters using affinity metadata and propagating and maintaining the affinity metadata in support of the described allocation. In order to maintain affinity metadata of the large file block cluster, affinity generation values stored on the large file block cluster are read and cached affinity generation values for each small file block cluster are read from an in-memory cache associated with the large file block cluster. When the stored affinity generation values and the cached affinity generation values do not match, affinity metadata from all the small file block clusters associated with the large file block cluster is used to update the affinity metadata of the large file block cluster and the associated cache.

Abstract: The disclosure provides an approach for performing a write to a storage system, the write having reduced latency due to parallelism of sub-steps of the write, and due to calculating a partial rather than a full checksum. The mechanism of the write has low file corruption of files on the storage system, due to journaling of the checksum.

Abstract: The disclosure herein describes managing a rate of processing unmap requests for a data storage volume. Unmap requests are received from a cluster of active hosts that are associated with the data storage volume. Latency data values of each active host are then accessed. A long-term cluster latency average value is calculated based on the accessed latency data values of all active hosts over a long-term time period and a short-term cluster latency average value is calculated based on the accessed latency data values of all active hosts over a short-term time period. An unmap rate adjustment value is calculated based on a difference between the long-term cluster latency average value and the short-term cluster latency average value. The rate of processing unmap requests for the data storage volume is adjusted based on the unmap rate adjustment value and the unmap requests are performed based on the adjusted rate.

Abstract: A method is provided for a computer to allocate a resource from a clustered file system (CFS) volume stored on one or more physical storage devices to a file. The CFS volume includes resources organized into resource clusters and the resource clusters make up regions. The method includes, for each region of resource clusters, determining a first count of resources allocated to the host computer and a second count of resources allocated to all other host computers, and calculating a region weight based on the first count and the second count. The method further includes sorting a list of the regions based on their region weights, selecting a region at or near the start of the list, and allocating the resource from a resource cluster in the selected region to the file.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for reclaiming one or more portions of storage resources in a computer system serving one or more virtual computing instances, where the storage resources in the computer system are organized in clusters of storage blocks. In one aspect, a method includes maintaining a respective block tracking value for each storage block that indicates whether a call to reclaim the storage block is outstanding; determining, from the block tracking values, a respective cluster priority value for each of the clusters based on a count of storage blocks in the respective cluster for which a call to reclaim is outstanding; and reclaiming a first portion of storage resources in the computer system in accordance with the cluster priority values.

Abstract: The disclosure provides an approach for zeroing allocated storage blocks of a file. The blocks are zeroed in the background, during a normal operation of a storage system, thus lowering the chance that the latency of a storage operation would be increased by the zeroing process. The approach also precludes a delay in being able to use the file, the delay caused by pre-zeroing the storage blocks prior to use of the file.

Abstract: The disclosure herein describes managing a rate of processing unmap requests for a data storage volume. Unmap requests are received from a cluster of active hosts that are associated with the data storage volume. Latency data values of each active host are then accessed. A long-term cluster latency average value is calculated based on the accessed latency data values of all active hosts over a long-term time period and a short-term cluster latency average value is calculated based on the accessed latency data values of all active hosts over a short-term time period. An unmap rate adjustment value is calculated based on a difference between the long-term cluster latency average value and the short-term cluster latency average value. The rate of processing unmap requests for the data storage volume is adjusted based on the unmap rate adjustment value and the unmap requests are performed based on the adjusted rate.

Abstract: The disclosure herein describes managing a rate of processing unmap requests for a data storage volume. Unmap requests are received from a cluster of active hosts that are associated with the data storage volume. Latency data values of each active host are then accessed. A long-term cluster latency average value is calculated based on the accessed latency data values of all active hosts over a long-term time period and a short-term cluster latency average value is calculated based on the accessed latency data values of all active hosts over a short-term time period. An unmap rate adjustment value is calculated based on a difference between the long-term cluster latency average value and the short-term cluster latency average value. The rate of processing unmap requests for the data storage volume is adjusted based on the unmap rate adjustment value and the unmap requests are performed based on the adjusted rate.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for reclaiming one or more portions of storage resources in a computer system serving one or more virtual computing instances, where the storage resources in the computer system are organized in clusters of storage blocks. In one aspect, a method includes maintaining a respective block tracking value for each storage block that indicates whether a call to reclaim the storage block is outstanding; determining, from the block tracking values, a respective cluster priority value for each of the clusters based on a count of storage blocks in the respective cluster for which a call to reclaim is outstanding; and reclaiming a first portion of storage resources in the computer system in accordance with the cluster priority values.

Abstract: A method is provided for a computer to allocate a file a resource from a clustered file system (CFS) volume stored on one or more physical storage devices. The CFS volume includes resources organized into resource clusters and the resource clusters make up regions. The method includes, for each region of resource clusters, determining a first count of resources allocated to the host computer and a second count of resources allocated to all other host computers, and calculating a region weight based on the first count and the second count. The method further includes sorting a list of the regions based on their region weights, selecting a region at or near the start of the list, and allocating the resource from a resource cluster in the selected region to the file.

Abstract: Techniques for decoupling the commit and replay of file system metadata updates in a clustered file system (CFS) are provided. In one embodiment, a CFS layer of a computer system can receive a file I/O operation from a client application, where the file I/O operation involves an update to a file system metadata resource maintained on persistent storage. In response, a journaling component of the CFS layer can execute a commit phase for committing the update to a journal on the persistent storage. The CFS layer can then return an acknowledgment to the client application indicating that the file I/O operation is complete, where the acknowledgement is returned prior to completion of a replay phase configured to propagate the update from the journal to one or more locations on the persistent storage where the file system metadata resource is actually stored.

Abstract: The disclosure provides an approach for zeroing allocated storage blocks of a file. The blocks are zeroed in the background, during a normal operation of a storage system, thus lowering the chance that the latency of a storage operation would be increased by the zeroing process. The approach also precludes a delay in being able to use the file, the delay caused by pre-zeroing the storage blocks prior to use of the file.

Abstract: The systems described herein are configured to enhance the efficiency of memory usage and access in a VM file system data store with respect to allocating memory in large and small file block clusters using affinity metadata and propagating and maintaining the affinity metadata in support of the described allocation. During storage of file data, an affinity identifier of the file data is determined. The affinity identifier is used to identify a large file block cluster and a small file block cluster within the identified large file block cluster. The file data is stored in the selected small file block cluster and affinity metadata of the selected small file block cluster is updated to reflect the storage of the file data.

Abstract: The disclosure provides an approach for performing a write to a storage system, the write having reduced latency due to parallelism of sub-steps of the write, and due to calculating a partial rather than a full checksum. The mechanism of the write has low file corruption of files on the storage system, due to journaling of the checksum.