Distributed storage system and operating method thereof
AssigneeSamsung Electronics Co., Ltd.
InventorMyoungwon Oh
Life of this patent
Estimating the term…
Abstract
Provided is an operating method of a distributed storage system, the distributed storage system including a first domain including a plurality of computing nodes and a plurality of storage nodes communicating with the plurality of computing nodes via a non-volatile memory express over fabrics protocol and a distribution manager. A second node communicates via a first node. The operating method includes detecting, occurrence of failure of the first node in the first domain; and performing, a recovery operation based on a first tree and a second tree in response to the occurrence of failure being detected by re-allocating, a third node of the first domain to the second node, whereby the second node is able to communicate via the third node.
Claims20
An operating method of a distributed storage system, the distributed storage system comprising a domain comprising a plurality of computing nodes and a plurality of storage nodes communicating with the plurality of computing nodes via a non-volatile memory express over fabrics (NVMe-oF) protocol and a distribution manager, the operating method comprising:
- detecting, by the distribution manager, an occurrence of a failure of a first node in the domain, wherein a second node communicates in the distributed storage system via the first node;
- and
- performing, by the distribution manager, a recovery operation based on a first tree and a second tree in response to the occurrence of the failure by re-allocating a third node of the domain to the second node, whereby the second node is able to communicate in the distributed storage system via the third node,
- wherein the first tree is generated based on a logical connection state between the plurality of computing nodes and the plurality of storage nodes and a storage capacity of the plurality of storage nodes, and
- the second tree is generated based on a physical connection state between the plurality of computing nodes and the plurality of storage nodes and a usage rate of the plurality of storage nodes.
A distributed storage system comprising:
- a plurality of computing nodes;
- a plurality of storage nodes configured to communicate with the plurality of computing nodes through a non-volatile memory express over fabrics (NVMe-oF) protocol;
- and
- a distribution manager configured to detect a first node in which a failure has occurred, wherein a second node communicates in the distributed storage system via the first node, and
- to perform a recovery operation based on a first tree and a second tree by re-allocating a third node of a fault domain to the second node, whereby the second node is able to communicate in the distributed storage system via the third node, and wherein the fault domain comprises the second node,
- wherein the first tree is generated based on a logical connection state between the plurality of computing nodes and the plurality of storage nodes and a storage capacity of the plurality of storage nodes, and the second tree is generated based on a physical connection state between the plurality of computing nodes and the plurality of storage nodes and a usage rate of the plurality of storage nodes.
An operating method of a distributed storage system, the distributed storage system comprising a domain comprising a plurality of computing nodes and a plurality of storage nodes communicating with the plurality of computing nodes via a non-volatile memory express over fabrics (NVMe-oF) protocol and a distribution manager, the operating method comprising:
- detecting, by the distribution manager, an occurrence of a failure of a first node, wherein a second node communicates in the distributed storage system via the first node;
- determining a fault domain comprising the second node as a storage node based on a first tree and determining a second tree based on the first node being a computing node;
- selecting, from the second tree, a peer storage node among a second plurality of storage nodes comprised in the fault domain;
- determining a third node that is a computing node, wherein the third node is allocated to the peer storage node;
- and
- re-allocating the third node to the second node, whereby the second node is able to communicate in the distributed storage system via the third node,
- wherein the first tree is generated based on a logical connection state between the plurality of computing nodes and the plurality of storage nodes and a storage capacity of the plurality of storage nodes, and the second tree is generated based on a physical connection state between the plurality of computing nodes and the plurality of storage nodes and a usage rate of the plurality of storage nodes.
Dependencies are inferred from the claim text.