Nutanix, Inc.

Emc Corporation

Andrey Alexseyevich Pyatkov

Karan Gupta

Parthasarathy Ramachandran

Pavan Kumar Konka

Artem Yevgenyevich Livshits

Abhishek Gupta

Brajesh Kumar Shrivastava

Prakash Narayanasamy

Arun Olappamanna Vasudevan

Angshuman Bezbaruah

Ethan Joseph Torretta

Nathan Edward Dire

Sydney Benjamin Melhuish

An indexing technique provides an index data structure for efficient retrieval of a snapshot from a long-term storage service (LTSS) of an archival storage system. The snapshot is generated from typed data of a logical entity, such as a virtual disk (vdisk). The data of the snapshot is replicated to a frontend data service of the LTSS sequentially and organized as one or more data objects for storage by a backend data service of LTSS in an object store of the archival storage system. Metadata associated with the snapshot (i.e., snapshot metadata) is recorded as a log and persistently stored on storage media local to the frontend data service. The snapshot metadata includes information describing the snapshot data, e.g., a logical offset range of a snapshot of the vdisk and, thus, is used to construct the index data structure. Notably, construction of the index data structure is deferred until after the entirety of the snapshot data has been replicated and received by the frontend data service.

Technique for efficiently indexing data of an archival storage system

A scalable multi-cluster resource sharing facility. A global witness process runs on a first computing platform that communicates over one or more networks to any number of nodes situated over two or more clusters. The global witness process listens on the network for occurrences of leadership and/or resource requests from nodes of different clusters. The global witness processes a request by retrieving a resource request and a respective last known state value, comparing the last known state value to a global stored state value, then storing a new state value when the respective last known state value is equal to the stored state value. Any number of contemporaneous requests can be processed by the global witness process, however only one request can be granted. The other requestors each receive a rejection of their resource request when their proffered last known state value is not equal to the stored state value.

Scalable leadership election in a multi-processing computing environment

A technique improves storage efficiency of an object store configured to maintain numerous snapshots for long-term storage in an archival storage system by efficiently determining data that is exclusively owned by an expiring snapshot to allow deletion of the expiring snapshot from the object store. The technique involves managing index data structures to enable efficient garbage collection across a very large number of data objects. When a snapshot expires, the technique obviates the need to scan the numerous snapshot data objects to determine which index structures are no longer needed and can be reclaimed (garbage collected). The technique is directed to management of underlying storage based on different sets of policies. When certain snapshots expire and are ready for deletion, the technique is directed to finding those data blocks that are no longer referenced (used) by any valid snapshots.

Garbage collection from archival of storage snapshots

Embodiments are directed towards managing an impact of a plurality of system and/or maintenance tasks distributed across a plurality of nodes. At least one initiator node may provide at least one disk request to at least one disk on at least one participant node. Each disk on each participant node with a workload, such as request wait time, above a maximum threshold and/or below a minimum threshold may be determined. If a disk is determined to have a workload above a maximum threshold and/or below a minimum threshold, then an impact of respective tasks on initiator nodes that provided requests to the determined disks may be reduced and/or increased, respectively. In some embodiments, an initiator node may reduce or increase an impact of a task by reducing or increasing a number of threads executing on that initiator node for that task.

Impact management of system tasks

Growing community of inventors

Seattle, WA, United States of America

Andrey Alexseyevich Pyatkov