Abstract: An example method includes identifying, based on a request to write data to a volume provided by storage nodes in a storage node cluster, a cluster hierarchy associated with the storage node cluster, the cluster hierarchy identifying storage characteristics of the storage nodes, wherein the cluster hierarchy is based on data characterizing performance and storage capabilities of the storage nodes; based on the request, selecting a storage node for storing the data in the volume, the selecting being based, at least in part, on the storage characteristics identified by the cluster hierarchy and on one or more data distribution parameters associated with the storage volume; and transmitting the data to the storage node.

Abstract: A single application may rely on data storage provided by a plurality of virtual data storage volumes mounted on one or more compute nodes. A crash-consistent snapshot of the state of the application may be generated. Each of the virtual data storage volumes may be quiesced prior to the capture of a plurality of volume-level snapshots, after which each volume is unquiesced. The crash-consistent snapshot may be used to restore the application to the state at the time the snapshot was captured.

Abstract: Clustered containerized applications are implemented with scalable provisioning. Methods include receiving a data storage request to store one or more data values in a storage volume implemented across a storage node cluster, the storage node cluster including a plurality of storage nodes including one or more storage devices having storage space allocated for storing data associated with the storage volume. Methods may further include identifying a cluster hierarchy associated with the storage node cluster, the cluster hierarchy identifying storage characteristics of the plurality of storage nodes, the cluster hierarchy also identifying physical location information for the plurality of storage nodes, the physical location information indicating node-to-node proximity on a network graph.

Abstract: A single application may rely on data storage provided by a plurality of virtual data storage volumes mounted on one or more compute nodes. A crash-consistent snapshot of the state of the application may be generated. Each of the virtual data storage volumes may be quiesced prior to the capture of a plurality of volume-level snapshots, after which each volume is unquiesced. The crash-consistent snapshot may be used to restore the application to the state at the time the snapshot was captured.

Abstract: Two servers implemented as containerized applications may manage the storage of data and the access of that data by compute jobs in a distributed system. A metadata server may distribute data on ingress and assign files to particular storage volumes. The metadata server may then provide a lookup function for files and be configured to distribute a file to other volumes when necessary. A job server may launch jobs as containerized applications and coordinate data access across jobs.

Abstract: A clustered storage system may include potentially many different nodes. A node may mount a virtual storage volume for the use of a container application at the node. The node may receive a request from a different node and respond by indicating whether the virtual storage volume is in active use. In this way, the clustered storage system may safely but forcibly unmount a virtual storage volume having a failed or hanging mount point so that the volume may be mounted on a different node.

Abstract: Distributed storage systems are implemented with application aware provisioning. Methods include receiving a volume provision request to allocate data storage space for a storage volume implemented across a storage node cluster. Methods also include selecting a storage pool among a set of candidate storage pools for provisioning the storage volume. Selecting the storage pool includes using application aware provisioning rules.

Abstract: A clustered storage system may include potentially many different nodes, each including a storage driver and a kernel module. A node may mount a virtual storage volume for the use of a container application at the node. The storage driver and kernel module may receive a request from a storage driver at a different node and respond by indicating whether the virtual storage volume is in active use. In this way, the clustered storage system may safely but forcibly unmount a virtual storage volume having a failed or hanging mount point so that the volume may be mounted on a different node.

Abstract: Distributed storage systems are implemented with rule-based provisioning. Methods include receiving a volume provision request to allocate data storage space for a storage volume implemented across a storage node cluster. Methods also include receiving one or more rules for provisioning the storage volume, each rule being based on labels for one or more storage pools. Methods may also include applying each rule to each candidate storage pool in a set of candidate storage pools to generate a rule score for each rule. Methods may further include adding rule scores for each candidate storage pool to generate a storage pool score for each storage pool. Methods may also include selecting a storage pool among the set of candidate storage pools for provisioning the storage volume, wherein selecting the storage pool includes comparing each storage pool score to determine which candidate storage pool has the highest storage pool score.

Abstract: A container orchestrator may communicate with a scheduler agent on a compute node to mount and/or unmount a virtual storage volume at the compute node. The scheduler agent may perform operations such as creating and/or deleting a mount path, mounting and/or unmounting the virtual storage volume, and/or marking the mount path as writable and/or read-only in order to eliminate cross-node race conditions.

Abstract: Techniques and mechanisms are provided for migrating data blocks around a cluster during node addition and node deletion. Migration requires no downtime, as a newly added node is immediately operational while the data blocks are being moved. Blockmap files and deduplication dictionaries need not be updated.

Abstract: Two servers implemented as containerized applications may manage the storage of data and the access of that data by compute jobs in a distributed system. A metadata server may distribute data on ingress and assign files to particular storage volumes. The metadata server may then provide a lookup function for files and be configured to distribute a file to other volumes when necessary. A job server may launch jobs as containerized applications and coordinate data access across jobs.

Abstract: Techniques and mechanisms described herein facilitate the execution of a software program container having a plurality of layers. Each layer may include a plurality of files. For instance, a base layer may include a version of an operating system, while an upper layer may include a software program configured to run within the operating system. Different layers may store, create, or modify the same file. However, the precedence of the version of the data for the file may be based on an ordering of the layers. For example, the version of the file in a higher layer should supersede the version of the file in a lower layer. In some implementations, a privileged storage container that runs atop a virtualization layer may manage storage resources for other containers. The privileged storage container may store each file in a software program container as a snapshotable object.

Abstract: Clustered containerized applications are implemented with scalable provisioning. Methods include receiving a data storage request to store one or more data values in a storage volume implemented across a storage node cluster, the storage node cluster including a plurality of storage nodes including one or more storage devices having storage space allocated for storing data associated with the storage volume. Methods may further include identifying a cluster hierarchy associated with the storage node cluster, the cluster hierarchy identifying storage characteristics of the plurality of storage nodes, the cluster hierarchy also identifying physical location information for the plurality of storage nodes, the physical location information indicating node-to-node proximity on a network graph.

Abstract: A clustered storage system may include potentially many different nodes, each including a storage driver and a scheduler agent. A node may mount a virtual storage volume for the use of a container application at the node. When a master node receives a request to instantiate a new containerized application, a scheduler at the master node may determine prioritization information for the new containerized application. The prioritization information may prioritize one or more nodes at which a virtualized storage volume employed by the application container instance is mounted.

Abstract: Two servers implemented as containerized applications may manage the storage of data and the access of that data by compute jobs in a distributed system. A metadata server may distribute data on ingress and assign files to particular storage volumes. The metadata server may then provide a lookup function for files and be configured to distribute a file to other volumes when necessary. A job server may launch jobs as containerized applications and coordinate data access across jobs.

Abstract: A clustered storage system may include potentially many different nodes, each including a storage driver and a kernel module. A node may mount a virtual storage volume for the use of a container application at the node. The storage driver and kernel module may receive a request from a storage driver at a different node and respond by indicating whether the virtual storage volume is in active use. In this way, the clustered storage system may safely but forcibly unmount a virtual storage volume having a failed or hanging mount point so that the volume may be mounted on a different node.

Abstract: Systems, devices, and methods are disclosed herein for containerized scalable storage applications. Methods may include instantiating an application instance based on a plurality of application instance parameters, the application instance being configured to utilize a plurality of storage volumes implemented in a storage cluster. Methods may also include enumerating a plurality of unattached storage volumes included in the cluster associated with the application instance, the plurality of unattached storage volumes having a plurality of underlying physical storage devices, and the plurality of unattached storage volumes being identified based on a plurality of application instance parameters. The methods may further include attaching at least some of the plurality of unattached storage volumes to the application instance, wherein the attaching enables the application instance to access data stored in the attached storage volumes.

Abstract: A single application may rely on data storage provided by a plurality of virtual data storage volumes mounted on one or more compute nodes. A crash-consistent snapshot of the state of the application may be generated. Each of the virtual data storage volumes may be quiesced prior to the capture of a plurality of volume-level snapshots, after which each volume is unquiesced. The crash-consistent snapshot may be used to restore the application to the state at the time the snapshot was captured.

Abstract: Systems, devices, and methods are disclosed herein for containerized scalable storage applications. Methods may include instantiating an application instance based on a plurality of application instance parameters, the application instance being configured to utilize a plurality of storage volumes implemented in a storage cluster. Methods may also include enumerating a plurality of unattached storage volumes included in the cluster associated with the application instance, the plurality of unattached storage volumes having a plurality of underlying physical storage devices, and the plurality of unattached storage volumes being identified based on a plurality of application instance parameters. The methods may further include attaching at least some of the plurality of unattached storage volumes to the application instance, wherein the attaching enables the application instance to access data stored in the attached storage volumes.