Vast Data Ltd.

Alex Turin

Asaf Levy

Renen Hallak

Yogev Vaknin

Alon Horev

Avi Goren

Yuval Mintz

Vladimir Zdornov

Ido Yellin

Jeff Denworth

Andrew Pernsteiner

A method for resource allocation in a storage system, the method may include determining to reduce a load imposed on a compute core of the storage system, wherein the compute core is responsible for executing (a) one or more background task shards, and (b) at least one non-sharded task for responding to an input/output (I/O) request sent to the storage system; and reducing the load imposed on the compute core by reallocating at least one background task shard of the one or more background task shards to another compute core of the storage system.

Resource allocation in a storage system

A method for resolving an erred input/output (IO) flow, the method may include (i) sending over a path a remote direct write request associated with a certain address range; wherein the path is formed between a compute node of a storage system to a storage drive of the storage system; (ii) receiving by the compute node an error message related to the remote direct write request; wherein the error message does not indicate whether an execution of the remote direct write request failed or is only temporarily delayed; (iii) responding by the compute node to the error message by (a) preventing from sending one or more IO requests through the path, (b) preventing from sending at least one IO requests aimed to the certain address range; and (c) requesting, using a management communication link, to force an execution of pending IO requests that are related to the path; and (iv) reuse the path, by the compute node, following an indication that there are no pending IO requests that are related to the path.

Resolving erred 10 flows

A system and method for prolonging lifespans of storage drives. The method includes determining an expected expiration time for each of a plurality of blocks, wherein each block includes data of a respective file, wherein the expected expiration of each block is determined based on a file type of the respective file; and writing a portion of data to at least one block of the plurality of blocks based on the expected expiration time for each block.

Techniques for prolonging lifespan of storage drives

A system and method for scalable storage. The method includes placing a lock on a portion of a storage node, wherein placing the lock further comprises replacing a first value stored in the storage node with a second value using an atomic operation, wherein the atomic operation replaces the first value with the second value when the first value indicates an empty lock status, wherein the second value indicates an active lock status; allocating a storage location in the storage node by updating metadata stored in the locked portion of the storage node when the lock has been placed; and releasing the lock, wherein releasing the lock further comprises replacing the second value with a third value, wherein the third value indicates the empty lock status.

Techniques for scalable storage without communication on the synchronous path

A distributed storage system. The distributed storage system includes a plurality of storage nodes (DNodes), wherein the DNodes are configured to store a plurality of elements and a plurality of data blocks, wherein each element is a persistent metadata structure, wherein at least one of the elements store at least one attribute, wherein the at least one attribute includes a plurality of pointers; and a plurality of compute nodes (CNodes), wherein each CNode has access to each of the DNodes, wherein each CNode is assigned to a respective at least one cluster of a plurality of clusters, wherein each CNode is configured to receive at least a portion of access commands directed to the respective at least one cluster and to execute the received access commands based on the plurality of elements.

Techniques for clustering compute nodes in distributed storage

A storage system. The storage system includes a plurality of storage nodes (DNodes), wherein the DNodes are configured to store a plurality of elements and a plurality of data blocks, wherein each element is a persistent metadata structure, wherein at least one of the elements store at least one attribute, wherein the at least one attribute includes a plurality of pointers, wherein at least one of the data blocks is stored in at least one of the elements when the at least one of the data blocks has a size below a threshold size; and a plurality of compute nodes (CNodes), wherein each CNode has access to each of the DNodes, wherein each CNode is configured to receive an access command and to execute the access command based on the elements.

Distributed scalable storage

Systems and methods for leader election. A disclosed method includes sending, by a first compute node of a plurality of compute nodes, a plurality of remote procedure calls (RPCs) to a plurality of storage boxes according to an order, wherein each of the plurality of RPCs causes a leader election algorithm to execute in one of the plurality of storage boxes; and updating a state of the first compute node to 'leader' when a result of executing the leader election algorithm for each of the plurality of RPCs indicates that the first compute node is elected as a leader node.

System and method for leader election in distributed storage systems

A system and method for prolonging lifespans of storage drives. The method includes writing at least a first portion of data to at least one high endurance drive; and writing at least a second portion of data to at least one low endurance drive, wherein a storage system includes a plurality of drives, wherein the plurality of drives includes the at least one high endurance drive and the at least one low endurance drive, wherein each high endurance drive has a longer total lifespan than each low endurance drive.

A storage system. The storage system includes a plurality of storage nodes (DNodes), wherein the DNodes are configured to store a plurality of elements and a plurality of data blocks, wherein each element is a persistent metadata structure, wherein at least one of the elements store at least one attribute, wherein the at least one attribute includes a plurality of pointers; and a plurality of compute nodes (CNodes), wherein each CNode has access to each of the DNodes, wherein each CNode is configured to receive an access command and to execute the access command based on the elements.

Growing community of inventors

Kiryat Ono, Israel

Alex Turin