Metadata management method for NAS global namespace design

• US 8,458,299 B2
• Filed: 03/17/2010
• Issued: 06/04/2013
• Est. Priority Date: 03/17/2010
• Status: Expired due to Fees

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Lin, Wujuan, Kodama, Shoji, Utsunomiya, Naoki, Kamei, Hitoshi
• Primary Examiner(s): Etienne, Ario
• Assistant Examiner(s): Williams, Clayton R
• Application Number:
  US12/726,337
• Publication Number:
  US 20110231524A1
• Time in Patent Office: 1,175 Days
• Field of Search: None
• US Class Current: 709/220
• CPC Class Codes: G06F 16/1824 implemented using Network-attached Storage [NAS] architecture distributed or networked storage systems G06F3/067; protocols for distributed storage of data in a network H04L67/1097
• CPC Class Codes: G06F 3/0635 by changing the path, e.g. traffic rerouting, path reconfiguration
• CPC Class Codes: G06F 3/0667 at data level, e.g. file, record or object virtualisation
• CPC Class Codes: G06F 3/067 Distributed or networked storage systems, e.g. storage area networks [SAN], network attached storage [NAS]

Abstract

A global namespace (GNS) is constructed across heterogeneous network-attached storage (NAS) devices by organizing metadata servers into a cluster layer having cluster nodes and a consistent hashing overlay layer having overlay nodes. The global namespace is distributed based on workload at the cluster and overlay nodes, and a portion of the GNS namespace may be migrated from one layer to the other. Cluster nodes may be added to the cluster layer, overlay nodes may be added to the consistent hashing overlay layer. Further, a cluster node may become an overlay node or an overlay node may become a cluster node. To access the global namespace, a request for namespace metadata managed by a cluster node is serviced by the cluster node, but if the requested namespace metadata have been migrated to the consistent hashing overlay layer, the cluster node will forward the request to the responsible overlay node for servicing.

Claims

1. A method of managing a global namespace (GNS) for a network system including a plurality of network-attached storage (NAS) devices connected together by a network, comprising the steps of:

organizing a plurality of metadata servers each into one of a cluster layer having a plurality of cluster nodes and a consistent hashing overlay layer having a plurality of overlay nodes,

constructing a GNS namespace, including steps of creating a GNS hierarchy above share folders of the NAS devices and constructing the GNS hierarchy under the share folders, wherein the creating and constructing steps are performed by a cluster node in the cluster layer,

synchronizing the constructed GNS namespace across all cluster nodes in the cluster layer,

setting first and second portions of the GNS namespace to be managed by the cluster layer and consistent hashing overlay layer, respectively, and

dynamically adjusting at least one of the first and second portions of the GNS namespace to be managed by the cluster layer and consistent hashing overlay layer, respectively, based on workload of the metadata servers,

wherein the dynamically adjusting step includes the steps of initiating, by a cluster node in the cluster layer, a process to add at least one metadata server to the cluster layer or to the consistent hashing overlay layer when the average total workload in the layer to which at least one metadata server is to be added, defined as the average of the number of namespace access requests in a predefined time period divided by the total number of namespace access requests than can be serviced during the predefined time period by the layer to which at least one metadata server is to be added, exceeds a predefined threshold;

wherein the dynamically adjusting step further includes the step of initialing, by a cluster node in the cluster layer, a process to shift part of the first portion to the second portion to be managed by the consistent hashing overlay layer, or part of the second portion to the first portion to be managed by the cluster layer, based on the workload of the metadata servers; and

wherein the dynamically adjusting step further includes the step of changing the metadata format from a first format required by the cluster layer to a second format required by the consistent hashing overlay layer when shifting part of the first portion to the second portion, or changing the metadata format from the second format to the first format when shifting part of the second portion to the first portion.

2. The method of managing a global namespace (GNS) according to claim 1, wherein the process to add at least one metadata server adds at least one cluster node as the metadata server to the cluster layer or at least one overlay node as the metadata server to the consistent hashing overlay layer.

3. The method of managing a global namespace (GNS) according to claim 1, wherein the process to add at least one metadata server migrates at least one cluster node as the at least one metadata server from the cluster layer to the consistent hashing overlay layer when the synchronization workload, defined as the number of namespace synchronization requests in a predefined time period divided by the total number of namespace access requests that can be serviced during the predefined time period by the cluster layer, exceeds a predefined threshold.

4. The method of managing a global namespace (GNS) according to claim 1, wherein the process to add at least one metadata server migrates at least one overlay node from the consistent hashing overlay layer to the cluster layer when the average total workload in the cluster layer, defined as the number of namespace access requests in a predefined time period divided by the total number of namespace access requests that can be serviced during the predefined time period by the cluster layer, is not greater than a predefined threshold.

5. The method of managing a global namespace (GNS) according to claim 1, wherein the initiating step is performed by a cluster node in the cluster layer and each overlay node in the consistent hashing overlay layer.

6. The method of managing a global namespace (GNS) according to claim 1, wherein the initiating step is performed by a cluster node in the cluster layer and each overlay node in the consistent hashing overlay layer.

7. The method of managing a global namespace (GNS) according to claim 1, further comprising the steps of:

accessing the global namespace, including steps of;

submitting, to a cluster node in the cluster layer, a request for access to the global namespace;

if namespace metadata of the request are managed by the cluster node to which the request is submitted, the cluster node services the request;

if namespace metadata of the request have been migrated from the cluster layer to the consistent hashing overlay layer, the cluster node forwards the request to the overlay node to which the namespace metadata was migrated at the consistent hashing overlay layer, and the overlay node to which the request is forwarded services the request.

8. The method of managing a global namespace (GNS) according to claim 7, wherein the node that services the request:

checks whether the request is a read request to read an existing GNS namespace, a create request to create a new GNS path entry, or a migrate request to migrate a file/directory from one NAS device to another, and if the request is a read request, a read-request process is invoked;

if the request is a create request, a create-request process is invoked; and

if the request is a migrate request, a migrate-request process is invoked;

in the read-request process, retrieves a requested file/directory from a location recorded in the namespace metadata; and

responds to the request with the retrieved file/directory;

in the create-request process, obtains a location of a GNS parent directory;

checks whether the parent directory is stored in a NAS share folder of a NAS device;

if the parent directory is stored in a NAS share folder, creates a file/directory under the parent directory in the NAS share folder;

if the parent directory is not stored in a NAS share folder, creates the file/directory in any of the NAS share folders; and

creates a GNS namespace entry with a location where the file/directory is created; and

in the migrate-request process, extracts a migration destination location for the migrate request;

retrieves a file/directory requested from a requesting location recorded in the namespace metadata;

stores the file/directory to the migration destination location;

updates the location of the migrated file/directory in the namespace metadata to be the migration destination location; and

removes the file/directory from the requesting location.

9. A method of managing a global namespace (GNS) for a network system including a plurality of network-attached storage (NAS) devices connected together by a network, comprising the steps of:

in an initialization phase, creating a cluster layer of cluster nodes, each cluster node having a GNS-node information table and a NAS information table, obtaining a node ID for each cluster node and entering the node ID of each cluster node into its GNS-node information table, and synchronizing the GNS-node information tables of the cluster nodes;

in a namespace construction phase, gathering information of NAS nodes into the NAS information tables, creating a GNS hierarchical namespace above NAS shares, constructing the GNS hierarchical namespace below the NAS shares, and synchronizing the NAS information tables of the cluster nodes;

in a workload monitoring phase, updating workload information in the GNS-node information tables, gathering workload information from a consistent hashing overlay layer, updating the GNS-node information table, and synchronizing the GNS-node information tables of the cluster nodes; and

in a namespace redistribution phase, when the average total workload of the cluster nodes is greater than a predefined first threshold, adding at least one new cluster node to the cluster layer;

when the average total workload of the consistent hashing overlay layer is greater than a predefined second threshold, adding at least one new overlay node to the consistent hashing overlay layer;

when the synchronization workload of a cluster node is greater than a predefined third threshold, migrating at least one cluster node to the consistent hashing overlay layer; and

when the total average workload in the cluster layer is not greater than a predefined fourth threshold, migrating at least one overlay node to the cluster layer.

10. The method of managing a global namespace (GNS) according to claim 9, wherein a cluster node in the cluster layer determines whether the synchronization workload of a cluster node is greater than the predefined third threshold and whether the total average workload in the cluster layer is not greater than the predefined fourth threshold.

11. The method of managing a global namespace (GNS) according to claim 9, wherein a cluster node in the cluster layer and each overlay node in the consistent hashing overlay layer determines whether the synchronization workload of a cluster node is greater than the predefined third threshold and whether the total average workload in the cluster layer is not greater than the predefined fourth threshold.

12. The method of managing a global namespace (GNS) according to claim 9, wherein the steps of migrating at least one cluster node from the cluster layer to the consistent hashing overlay layer and migrating at least one overlay node from the consistent hashing layer to the cluster layer further include the steps, of changing the metadata format from a first format required by the cluster layer to a second format required by the consistent hashing overlay layer and changing the metadata format from the second format to the first format, respectively.

13. A network system including a plurality of cluster nodes in a cluster layer and a plurality of overlay nodes in a consistent hashing overlay layer, the cluster layer and consistent hashing overlay layer being connected together and to a plurality of NAS devices via a network, wherein:

each cluster node and each overlay node includes a processor, a network interface via which the cluster node is coupled to the network, an NFS/CIFS protocol module, a storage management module, a storage interface, a system memory, and a system bus via which the processor is coupled to each of the network interface, the NFS/CIFS protocol module, the storage management module, the storage interface, and the system memory,

wherein the processor executes;

an initialization program, in an initialization phase, creating a cluster layer of cluster nodes, each cluster node having a GNS-node information table and a NAS information table, obtaining a node ID for each cluster node and entering the node ID of each cluster node into its GNS-node information table, and synchronizing the GNS-node information tables of the cluster nodes;

a namespace construction program, in a namespace construction phase, gathering information of NAS nodes into the NAS information tables, creating a GNS hierarchical namespace above NAS shares, constructing the GNS hierarchical namespace below the NAS shares, and synchronizing the NAS information tables of the cluster nodes;

a workload monitoring program, in a workload monitoring phase, updating workload information in the GNS-node information tables, gathering workload information from a consistent hashing overlay layer, updating the GNS-node information table, and synchronizing the GNS-node information tables of the cluster nodes; and

a namespace redistribution program, in a namespace redistribution phase, when the average total workload of the cluster nodes is greater than a predefined first threshold, adding at least one new cluster node to the cluster layer, when the average total workload of the consistent hashing overlay layer is greater than a predefined second threshold, adding at least one new overlay node to the consistent hashing overlay layer;

when the synchronization workload of a cluster node is greater than a predefined third threshold, migrating at least one cluster node to the consistent hashing overlay layer; and

when the total average workload in the cluster layer is not greater than a predefined fourth threshold, migrating at least one overlay node to the cluster layer.

14. The network system according to claim 13, wherein a cluster node in the cluster layer determines whether the synchronization workload of a cluster node is greater than the predefined third threshold and whether the total average workload in the cluster layer is not greater than the predefined fourth threshold.

15. The network system according to claim 13, wherein a cluster node in the cluster layer and each overlay node in the consistent hashing overlay layer determines whether the synchronization workload of a cluster node is greater than the predefined third threshold and whether the total average workload in the cluster layer is not greater than the predefined fourth threshold.

16. The network system according to claim 13, wherein migrating at least one cluster node to the consistent hashing overlay layer and migrating at least one overlay node to the cluster layer further include changing the metadata format from a first format required by the cluster layer to a second format required by the consistent hashing overlay layer and changing the metadata format from the second format to the first format, respectively.