Method for data maintenance

Abstract

A method for data storage implemented in a data storage system is disclosed. Data storage nodes may be interconnected by means of a communications network. The method may include sending a request for a first data item to a plurality of storage nodes. The first data item may include a reference to a second data item stored in the storage system. The method may include receiving the first data item from at least one storage node, and sending a request for the second data item to the plurality of storage nodes based on the reference included in the first data item.

Description

TECHNICAL FIELD

The present disclosure relates to an apparatus and method for accessing, writing and deleting data in a data storage system comprising a plurality of data storage nodes, the methods may be employed in a server and/or in a storage node in the data storage system. The disclosure further relates to storage nodes or servers that may be capable of carrying out such methods.

BACKGROUND

Such a method is disclosed in US Patent Publication No. 2005/0246393 A1. This method is disclosed for a system that may use a plurality of storage centers at geographically disparate locations. Distributed object storage managers may be included to maintain information regarding stored data. One problem associated with such a system is how to accomplish simple and yet robust and reliable maintenance of data.

SUMMARY OF THE INVENTION

A method to maintain data in a data storage system comprising data storage nodes interconnected by means of a communications network is disclosed. The method may include sending a request for a first data item to a plurality of storage nodes. The first data item may include a reference to a second data item stored in the storage system. The method may also include receiving the first data item from at least one storage node, and/or sending a request for the second data item to the plurality of storage nodes, for example based on the reference included in the first data item.

In an embodiment, a virtual directory structure may be implemented in a storage system where the files are stored in an unstructured manner.

The request for the first and second data entities may be sent by multicast.

For example, by utilizing multicast many storage nodes may be easily accessed.

The first and second data items may be identified by a first and second unique key.

In an embodiment, the files in the storage system may be accessed directly irrespective of their location in the system.

The first and second unique key may include a cluster address pointing out a subset of said storage nodes within the system, and/or a data item identifier identifying a data item within the subset of storage nodes. A reference to a data item may include the unique key.

In an embodiment, the method may be implemented in large storage systems, for example a storage system which include hundreds or thousands of storage nodes.

The method may comprise sending the first and second request from an application programming interface, API.

For example, by utilizing a common API to access the storage nodes, the method may be easily implemented on many different platforms.

The API may be implemented on a server in communication with the storage nodes.

In an embodiment, the method may be implemented on a dedicated device provided e.g. by a third party which may not be responsible for the maintenance of the storage nodes.

The API may be implemented in a storage node.

In an example embodiment, implementing the API in a storage node may allow the number of access points into the storage system may be increased.

The method may comprise the API retrieving the unique key, identifying the second data item, from the received first data item.

For example, the unique identifier for the second data item may be easily retrievable if an indication of the unique identifier is included in the first data item.

The method may comprise receiving the key, identifying the first data item, in the API.

In an embodiment, one or more directory structures may be implemented simultaneously.

The second data item may include a reference to third data item.

For example, in an embodiment directory structures with multiple levels may be implemented.

The second data item may include payload data, such as an image.

For example, data files with payload data may be part of the directory structure (e.g. stored in sub folders).

The first data item may be sent by unicast.

For example, by utilizing unicast, the data item may be transferred in a bandwidth-effective manner.

According to an embodiment, a method for maintaining data in a data storage system including data storage nodes interconnected by means of a communications network may be implemented in a server and/or a data storage node. The method may include storing a first data item in at least one storage node. The method may also include updating a second data item, stored in at least one storage node. For example, the second data item may be updated by adding a reference to the first data item in the second data item. Updating the second data item may include sending a request to at least one storage node that stores a copy of the second data item. The request may instruct and/or request that the at least one storage node add a reference to the first data item to the second data item.

In an embodiment, new items may easily be added to the directory structure, for example by adding references to the new items to other items in the directory structure.

According to an embodiment, a method for maintaining data may be implemented a server or data storage node included in a data storage system. The data storage nodes may be interconnected by means of a communications network. The method may include deleting a first data item stored in at least one storage node. The method may also include updating a second data item, stored in at least one storage node, by deleting a reference to the second data item in the second data item.

In an example embodiment, items in the directory structure may easily be deleted, for example by deleting references to the items.

According an embodiment, a data storage system may include data storage nodes interconnected by means of a communications network. A server or node may include an application programming interface, API, and may be configured to send a request for a first data item to a plurality of storage nodes. The first data item may include a reference to a second data item stored in the storage system. At least one storage node may be configured to send the first data item to the API. The API and/or storage node or sever may be further configured to send a request for a second data item to the plurality of storage nodes based on the reference included in the first data item.

For example, a virtual directory structure may be implemented in a storage system where the files are stored in an unstructured manner.

Other objectives, features and advantages of the disclosed embodiments may appear from the following detailed disclosure, from the attached claims as well as from the drawings.

Generally, all terms used in the claims are to be interpreted according to their ordinary meaning in the technical field, unless explicitly defined otherwise herein. All references to “a/an/the [element, device, component, means, step, etc.]” are to be interpreted openly as referring to at least one instance of said element, device, component, means, step, etc., unless explicitly stated otherwise. The steps of any method disclosed herein do not have to be performed in the exact order disclosed, unless explicitly stated.

BRIEF DESCRIPTION OF THE DRAWINGS

The above, as well as additional objects, features and advantages of the disclosed embodiments, may be better understood through the following illustrative and non-limiting detailed description, with reference to the appended drawings, where the same reference numerals may be used for similar elements.

FIG. 1 is a schematic view of an example storage system.

FIG. 2 is an example schematic block diagram of a number of data items stored in the storage system.

FIG. 3 is a schematic block diagram of an example data item identifier.

FIG. 4 is a schematic block diagram of an example method for retrieving data.

FIG. 5a-c are illustrations of example communications between different entities in the storage system.

FIG. 6 is a schematic block diagram of an example method for storing data.

FIG. 7 is a schematic block diagram of an example method for deleting data.

DETAILED DESCRIPTION

Detailed embodiments of the disclosed methods and systems may be described with reference to the drawings. The present disclosure is related to a distributed data storage system comprising a plurality of storage nodes. An example structure of the system and the context in which it may be used is outlined in FIG. 1.

A user computer 1 may access, for example via the Internet 3, an application 5 running on a server 7. The user context, as illustrated here, may be therefore a client-server configuration. However, it should be noted that the data storage system to be disclosed may be useful also in other configurations, for example utilizing other communication methods.

In the illustrated case, two applications 5, 9 may run on the server 7. Of course however, any number of applications may be running on server 7. Each application may have an API (Application Programming Interface) 11 which may provide an interface in relation to the distributed data storage system 13 and may support requests, typically write and read requests, from the applications running on the server. Data may be read and written to the storage system using the methods described in detail in U.S. patent application Ser. No. 13/125,524, filed Apr. 21, 2011, the contents of which are hereby incorporated by reference herein. Methods of reading and writing of data therefore may not be further elaborated upon in detail herein. From an application's point of view, reading or writing information from/to the data storage system 13 may appear to be the same as using any other type of storage solution, for instance a file server or a hard drive.

Each API 11 may communicate with storage nodes 15 in the data storage system 13, and the storage nodes may communicate with each other. Alternatively, or additionally, one or more of the storage nodes 15 may include an API 23 for supporting requests as disclosed above. These communications may be based on TCP (Transmission Control Protocol) and UDP (User Datagram Protocol). Other communication protocols may also be utilized.

The components of the distributed data storage system may be the storage nodes 15 and the APIs 11 in the server 7 which may access the storage nodes 15. The present disclosure may be described in relation to methods carried out in the server 7 and in the storage nodes 15. Those methods may be primarily be embodied as combination software/hardware implementations which are executed on the server and the storage nodes, respectively. The operations of the server and/or storage nodes may together determine the operation and the properties of the overall distributed data storage system.

Although in FIG. 1 the server 7 is illustrated as a member of the storage system 13 which is separate from the storage nodes 15 it should be noted that the server 7 may be a storage node including server functionality.

The storage node 15 may typically be embodied by a file server which is provided with a number of functional blocks. The storage node may thus include a storage medium 17, which for example may include a number of internal (e.g. connected via Integrated Drive Electronics (IDE), Serial Advanced Technology Attachment (SATA), and/or the like) or external hard drives (e.g. connected via Universal Serial Bus (USB), Firewire, Bluetooth, and/or the like), optionally configured as a RAID (Redundant Array of Independent Disk) system. Other types of storage media are however conceivable as well.

Each storage node 15 may contain a node list including the IP addresses of all storage nodes in its set or group of storage nodes. The number of storage nodes in a group may vary from a few to hundreds or thousands of storage nodes.

The storage medium 17 may store one or more data items 19, 21 in the form of collection objects 19 or payload data in the form of data files 21. A collection object 19 may include a set of references. A reference may be a reference to one or more data files stored in the storage system, e.g. data files 21. A reference may also be a reference to another collection object 19 stored in the storage system. A reference may include a pointer (e.g. a memory address) to a storage location of a storage node 15. A reference may include an identifier of the collection object or data file referred to.

As will be disclosed in more detail below, the collection object 19 may be used for implementing a structured layer in the storage system. Data files 21 referenced in the collection object 19 may in such an implementation represent data files stored in the structure. Additional collection objects 19 referenced in the collection object 19 may in such an implementation represent subdirectories stored in the directory.

A collection object 19 may be embodied as a data object having a predetermined format. The data object may be a special file in the file system of the storage medium 17 in the sense that it may be a binary file to be interpreted by the API. In an example, the data object may be a standard data file in the file system of the storage medium 17; the data object may e.g. be a plain text file indicating the referenced collection objects 19 and/or data files 21. A data object may be readable using the same routines of the file system as the data files 21.

FIG. 2 schematically illustrates a collection object 19a according to one embodiment. The collection object 19a may have an associated collection object identifier 20a. The identifier 20a may for instance be a Universally Unique Identifier (UUID). The collection object identifier 20a may be included in a header of the collection object 19a. However, the collection object identifier 20a may be stored in a register maintained at the storage node 15, for example rather than being included in the collection object 19a. In an example, the UUID and/or the register maintained at storage node 15 may associate the collection object 19a with the collection object identifier 20a, e.g. by pointing out the memory address where the collection object 19a is to be found. Thus, the collection object 19a may form a first data item which is identified by a first unique key.

The collection object 19a may include a field 22a with an identifier 20b of another collection object 19b, e.g. in the form of a string. The collection object 19a may include a reference to the collection object 19b. The collection object 19b may be stored on the same storage node as the collection object 19a or on another storage node than the collection object 19a. The storage system may use the identifier 20b in the field 22a to locate and access the collection object 19b. Thus, the collection object 19b may form a second data item which is identified by a second unique key.

In one embodiment, in order to implement large storage systems spanning over multiple networks, the data item identifiers 20a-d may include two data elements. With reference to FIG. 3, the first data element 30 may be a cluster ID 31 that may identify the cluster where the data item (collection object 19a-c or data file 21a) is located. The cluster address may be a multicast address 32. The multicast address 32 may be utilized by the API to send a request for a data item to a specific cluster. The second data element 33 may be a data item ID 34 formed by a unique number 35 that identifies the data item 19a-d inside the cluster. The unique number 35 may be a number of with a defined length, e.g. 128 bits, or the length may vary. The unique number 35 may include a large number of bits, enabling a large number of data items to be uniquely identified within the cluster. By this arrangement a collection element in one cluster may reference another collection element or data file in another cluster. In other words, the first and second unique key may include a cluster address pointing out a subset of the storage nodes within the system, and a data item identifier identifying a data item within the subset of storage nodes.

Referring back to FIGS. 1 and 2, the server 7 may for example include a register indicating a storage node 15 storing the collection object (e.g. collection object 19a) associated with a specific identifier (e.g. identifier 20a). In another example, the collection object 19a may be located using the read method disclosed in U.S. patent application Ser. No. 13/125,524. Briefly, according to this read method the server 7 or a storage node 15 may send a multicast message to the plurality of storage nodes 15. The multicast message may include the identifier 20a of the desired collection object 19a. Each storage node 15, in response to receiving the multicast message, may scan its storage medium 17 for a collection object having said identifier. If found, the storage node 15 may respond and indicate that it stores the object sought-after to the originator of the multicast message. The collection object 19a may then be accessed by means of a unicast request sent to a responding storage node 15 storing the collection object 19a.

According to the present embodiment, multicast communication may be used to communicate simultaneously with a plurality of storage nodes. By a multicast or IP multicast is here meant a point-to-multipoint communication which may be accomplished by sending a message to an IP address which may be reserved for multicast applications. For example, a message, for example a request, may be sent to such an IP address (e.g. 244.0.0.1), and a number of recipient servers may be registered as subscribers to that IP address. Each of the recipient servers may have its own IP address. When a switch in the network receives the message directed to 244.0.0.1, the switch may forward the message to the IP addresses of each server registered as a subscriber.

In principle, a single server may be registered as a subscriber to a multicast address, in which case a point-to-point, communication may be achieved. However, in the context of this disclosure, such a communication may nevertheless be considered a multicast communication since a multicast scheme is employed.

According to the present embodiment, unicast communication may refer to a communication with a single recipient. A unicast communication may be initiated by a party of the network and may be directed to a single specific recipient.

In addition to the collection object 19a, the collection object 19b may include a field 22b with an identifier 20c of a third collection object 19c. The collection object 19c may include a field 22c with an identifier 20d of a data file 21a. In other words, anyone of the collection objects 19a-c (or, for example, each of the collection objects 19a-c) may represent a second data item including a reference to third data item, and the data file 21a may represent a second data item including payload data, for example an image.

By appointing the collection object 19a as a root collection object, the collection object 19a may represent a root directory 19a of the storage system. Analogously the collection object 19b may represent a subdirectory 19b of the root directory 19a. The collection object 19c may represent a subdirectory of the subdirectory 19b. The data file 21a may represent a data file stored in the subdirectory 19c. The collection objects 19a-c may thus define a hierarchical storage structure. The structure may be referred to as a directory tree.

With reference to FIGS. 4 and 5a-c, a method for parsing a directory structure in order to access a file 19, 21 stored in a storage node 15 may be disclosed.

The starting point of the directory structure may be a predefined root key. For example, any of the storage nodes 15 may include a root key. This key may be stored outside of the storage cluster and may be used to identify the first data item (e.g. collection object 19a) in the directory structure. A storage cluster may have multiple root keys that enable the user to have multiple individual directory structures stored within the same storage cluster. The directory structures may span over several storage clusters. The root key may be stored together with external information describing the directory structure stored within the cluster.

In block 40, the server 7 may receive the root key, which may identify the first data item 19, 21 and may pass the unique identifier to identify the file within the storage system to the API 11. In an example, the API 23 may be implemented in a storage node 15, wherein the root key may be received in the storage node 15 rather than in the server 7.

In block 41, the API 11 in the server 7 may multicast a request for the data item (e.g. the collection object 19a) identified by the root key to storage nodes 15a-e in the storage system, or to a subset of the nodes. For example, the multicast message may be sent to a specific cluster, for example using the data item identifier configuration disclosed in relation to FIG. 3. According to one embodiment the data item (e.g. the collection object 19a) identified by the root key may be a special data item in the sense that it may include additional metadata which may be used by the system. Examples of such data may be information regarding access permissions to the items in the directory structure, information where to store certain data items (e.g. on a storage node with quick access such as an Solid State Drive (SSD)), and the like.

In block 42, storage node 15a-e, in response to receiving the multicast message, may scan their respective storage mediums 17 in an attempt to locate the data item identified by the data item ID 34 in the root key.

For purposes of illustration, it may be assumed in this example that nodes 15b and 15e locate the data item identified by the data item ID 34. In block 43, the nodes 15b, 15e that find the data item may reply with information about what other nodes 15b, 15d, 15e may contain the data item and the current execution load (e.g. how busy the nodes are, how many request the nodes received, how much free space is on the node, etc.) in the node 15b, 15e. The requested data item may be stored in a plurality of storage nodes 15b, 15d, 15e, wherein the API may collect the information received from the nodes 15b, 15d, 15e and may wait until it has received answers from more than 50% of the listed storage nodes 15b, 15d, 15e that contains the data item before it may make a decision on which one to select for the retrieval of the data item. The decision may be based on which node that has the lowest execution load.

In block 44 the API 11 may send a unicast request for the specific file to the chosen storage node. In this example, for purposes of illustration it may be assumed store node 15b is chosen. API 11 may retrieve the data item from the storage node 15b. The API 11 may maintain a list of all storage nodes 15b, 15d, 15e that store copies of the file in the event of a read or communication failure with the selected node 15b. If an error occurs the API 11 may transparently select the next best node in the list and continues the read operation.

In block 45 the API may interpret the content of the retrieved data item. If the directory structure comprises additional levels the retrieved data item may be a collection object 19b. If so, the API 11 may read the field 22b which may include an identifier 20b referring to another collection object 19c in the directory structure. For example, the API may retrieve the unique key, i.e. the identifier 20b, identifying the second data item, e.g. the collection object 19b, from the received first data item, e.g. the collection object 19a. The process may then return to block 41 and may continue parsing the directory structure. Thus, both the first and second request may be sent from an application programming interface, API. The process may continue until the last object in the directory structure has been identified and retrieved, e.g. the data file 21a whereupon the process may end at 46. In another example, the API 11 may send an update request to the identified object, e.g. a command to alter or concatenate data in the data item corresponding to the object in the directory structure.

As an example, it may be that the data file 21 is located in the root of the directory structure. In such a case the process may be looped a single time, since the first retrieved collection object 19a may contain a reference to the data file 21a. It is emphasized that the retrieved collection object in addition to including the reference to the data file 21a may also include references to other data items, such as collection object 19b.

Thus, according to the above, a method may implemented in a data storage system including data storage nodes interconnected by means of a communications network for accessing file. The method may include sending a request for a first data item 19, 21, (e.g. collection object 19a), to a plurality of storage nodes 15a-e. The first data item may include a reference to a second data item (e.g. data file 21a or collection object 19b), stored in the storage system. The method may include receiving the first data item from at least one storage node 15b, and sending a request for the second data item to the plurality of storage nodes 15a-e based on the reference included in the first data item.

As an illustrative example, with reference to FIG. 2, the API may recursively read and interpret referenced envelopes to resolve a path in a directory structure. For example, the API may identify an unstructured key that represents a file in the structured path. For example, a user accessing the storage system may want to resolve the path:

“/Documents/Sample_Pictures/Blue_Hills.jpg”.

In FIG. 2, collection object 19a may represent the root key “/” (identified by the unique key 20a) and the identifier 22a may include a reference to collection object 19b representing the folder “Documents/” (identified by the unique key 20b). The identifier 22b in the collection object 19b may include a reference to collection object 19c representing the folder “Sample_Pictures/”. Finally, the identifier 22c in the collection object 19c may include a reference to the data file 21a comprising the payload data for the file “Blue_Hills.jpg”. Thus, by recursively reading the references in the collection objects a virtual file structure may be created in an unstructured storage system.

With reference to FIGS. 6 and 5a-c, a method for parsing a directory structure in order to create a file 19, 21 in a storage node 15 is disclosed.

similarly to the system discloses in FIG. 4, the starting point of the directory structure is a predefined root key. The root key may be an arbitrary key, and there may be many root keys throughout the system. This key may be stored outside of the storage cluster and may be used to identify the first data item (e.g. collection object 19a) in the directory structure.

In block 60, the server 7 may receive the root key, and may pass the unique identifier to identify the file within the storage system to the API.

In block 61, the API 11 may resolve the path to the desired data item according to the method above.

In block 63, the API 11 in the server 7 may multicast a request for storing the data item (e.g. the collection object 19c) including the identifier to all storage nodes 15a-e in the storage system, or to a subset of the nodes, e.g. within a specific cluster, for example using the data item identifier configuration disclosed in relation to FIG. 3.

In block 63, storage nodes 15a-e, in response to receiving the multicast message, may verify that the data item ID 34 is not already in use.

In block 64, a storage node which 15a-e that fails to find an existing file with that specific identifier may reply with an acknowledgment that may indicate: free storage space on the storage node, an indication of the age of the hardware that the storage node is running on, current CPU load and/or the geographical position of the storage node 15a-e in the form of latitude, longitude and altitude or the like.

In block 65 the API 11 may select three storage nodes (e.g. storage nodes 15a, 15b and 15e) based on the data returned from storage nodes that responded to the multicast request. When the three most suitable nodes have been selected the API 11 may send a request to the three nodes simultaneously to store the data item. If an error occurs during the transfer of the data item to one of the selected nodes 15a, 15b, 15e, the operation continues, for example as long as more than 50% of the selected nodes are operational.

In block 66 the identifier field 22b in the data item one level higher up in the directory structure (e.g. the first data item—the collection object 19b) may be updated with a reference to the stored data item (e.g. the collection object 19c), either by retrieving the first data item according to the read method according to the above or by accessing the first data item directly, for example if the server has cached the identifier of the first data item.

In order to increase the data integrity in the system the method above may be supplemented with the act of, prior to storing the data item, retrieving the first data item in case the communication with all storage nodes should be lost after the data item has been stored but before the first data item is updated. By this procedure the API may resume the update procedure once the communication with the storage nodes is resumed.

Thus, according to the above, a method may be implemented in various devices within a data storage system including data storage nodes interconnected by means of a communications network. The method may include storing a first data item in at least one storage node, and updating a second data item, stored in at least one storage node, by adding a reference to the first data item in the second data item.

With reference to FIGS. 7 and 5a-c, a method for parsing a directory structure in order to delete a file 19, 21 in a storage node 15 is disclosed.

In similarity to the disclosure in relation to FIG. 4, the starting point of the directory structure may be a predefined but arbitrary root key. This key may be stored outside of the storage cluster and may be used to identify the first data item (e.g. collection object 19a) in the directory structure.

In block 70, the server 7 may receive the root key and may pass the unique identifier to identify the file within the storage system to the API.

In block 71, the API 11 may resolve the path to the desired data item according to the method above.

In block 72, the API 11 in the server 7 may multicast a query regarding the location of the data item (e.g. the collection object 19c) including the identifier to storage nodes 15a-e in the storage system, or to a subset of the nodes, e.g. within a specific cluster, for example using the data item identifier configuration disclosed in relation to FIG. 3.

In block 73, storage nodes 15a-e, in response to receiving the multicast message, may scan their respective storage medium 17 to locate data item identified by the data item ID 34.

In block 74, the nodes that locate the data item may reply with information regarding other nodes that may store the data item and the current execution load in the node. The requested data item may be stored in a plurality of storage nodes. The API may collect the information received from the nodes and may wait until it has received answers from more than 50% of the listed storage nodes that contains the data item before making a decision on which nodes to select for the deletion of the data item.

In block 75 the API 11 may send a unicast request to delete the specific file (e.g. the collection object 19c) to the chosen storage nodes.

In block 76 the identifier field 22b in a data item one level higher in the directory structure (e.g. the collection object 19b) may be updated by deleting the reference to the deleted data item (e.g. the collection object 19c). The update may occur by retrieving the first data item according to the read method described above and/or by accessing the first data item directly, for example if the server cached the identifier of the first data item. In the case where the data item to be deleted is located a number of levels down in the directory structure, the delete operation may be expressed as the method disclosed in relation to FIG. 4 with the addition of i) deleting the first data item, and ii) updating the second data item by deleting the reference to the first data item.

Thus, a data deletion method may be implemented in a data storage system including data storage nodes interconnected by means of a communications network. The method may include deleting a first data item stored in at least one storage node. The method may also include updating a second data item, stored in at least one storage node, by deleting a reference to the first data item in the second data item.

The collection objects 19 may be handle and maintained in manner similar to data files. This may allow the data to be stored in a flat storage structure, e.g. without any subdirectories or within a single directory. A virtual hierarchical storage structure may be created by adding collection objects 19 including references to other collection objects 19 and/or data files 21. It even allows the same data to be organized in several different virtual hierarchical storage structures by using different sets of collection objects 19.

For data security reasons, some or all information stored in the storage system (e.g. the collection objects 19 and/or the data files 21) may be stored redundantly in the storage system. The collection objects 19a-c and the data file 21a may be stored at two or more storage nodes 15. Each instance of a collection object or data file may be associated with the same identifier. In such a case, the above-described read method may result in a response from each storage node storing the collection object. A redundantly stored collection object may thus be retrieved from either one of or all of the storage nodes storing the collection object.

Several embodiments which demonstrate the disclosed method and system have been described. However, as is readily appreciated by a person skilled in the art, other embodiments in addition to the ones disclosed above are equally possible that are in accordance with the methods and products described herein. The foregoing examples are not meant to be limiting, and the scope of the protection is to be defined by the appended patent claims.

Claims

1. A data storage node comprising a processor configured to: identify a first data item in a data storage system, wherein the first data item comprises a first collection object, a collection object is a type of file utilized to implement a distributed directory structure, the first collection object comprises one or more references to one or more other data items in the data storage system, and the one or more references comprise a reference to a second data item stored in the data storage system and a reference to a data file, the data file comprising payload data;determine a multicast address used to request the second data item in a cluster of the data storage system based on the reference to the second data item included in the first data item, the multicast address corresponding to a cluster address of the cluster;send a request for the second data item to the multicast address determined from the reference to the second data item included in the first data item; andreceive the second data item based on the request.

2. The data storage node as in claim 1, wherein the second data item corresponds to a second collection object, the first collection object corresponds to a directory of a distributed file system, and the second collection object corresponds to a first sub-directory of the directory.

3. The data storage node as in claim 2, wherein the processor is configured to add a new data file to the first sub-directory corresponding to the second collection object by adding a reference to the new data file to the second collection object.

4. The data storage node as in claim 2, wherein the processor is configured to add a second sub-directory to the directory corresponding to the first collection object by adding a reference to a third collection object to the first collection object, the third collection object corresponding to the second sub-directory.

5. The data storage node as in claim 2, wherein the processor is configured to delete the first sub-directory corresponding to the second collection object by removing the reference to the second collection object from the first collection object.

6. A server comprising a processor configured to: identify a first data item in a data storage system, wherein the first data item comprises a first collection object, a collection object is a type of file utilized to implement a distributed directory structure, the first collection object comprises one or more references to one or more other data items in the data storage system, and the one or more references comprise a reference to a second data item stored in the data storage system and a reference to a data file, the data file comprising payload data;determine a multicast address used to request the second data item in a cluster of the data storage system based on the reference to the second data item included in the first data item, the multicast address corresponding to a cluster address of the cluster;send a request for the second data item to the multicast address determined from the reference to the second data item included in the first data item; andreceive the second data item based on the request.

7. The server as in claim 6, wherein the second data item corresponds to a second collection object, the first collection object corresponds to a directory of a distributed file system, and the second collection object corresponds to a first sub-directory of the directory.

8. The server as in claim 7, wherein the processor is configured to add a new data file to the first sub-directory corresponding to the second collection object by adding a reference to the new data file to the second collection object.

9. The server as in claim 7, wherein the processor is configured to add a second sub-directory to the directory corresponding to the first collection object by adding a reference to a third collection object to the first collection object, the third collection object corresponding to the second sub-directory.

10. The server as in claim 7, wherein the processor is configured to delete the first sub-directory corresponding to the second collection object by removing the reference to the second collection object from the first collection object.

11. A method for implementing a distributed directory structure in a data storage system, the method comprising: identifying a first data item in a data storage system, wherein the first data item comprises a first collection object, a collection object is a type of file utilized to implement the distributed directory structure, the first collection object comprises one or more references to one or more other data items in the data storage system, and the one or more references comprise a reference to a second data item stored in the data storage system and a reference to a data file, the data file comprising payload data;determining a multicast address used to request the second data item in a cluster of the data storage system based on the reference to the second data item included in the first data item the multicast address corresponding to a cluster address of the cluster;sending a request for the second data item to the multicast address determined from the reference to the second data item included in the first data item; andreceiving the second data item based on the request.

12. The method as in claim 11, wherein multiple copies of the second data item are received based on the request, each of the multiple copies being received from a different data storage node of the cluster.

13. The method as in claim 11, wherein the first data item and the second data item are stored in different clusters of data storage nodes.

14. The method as in claim 11, wherein the second data item comprises a second collection object, the second collection object comprises at least one reference to at least one other data item in the data storage system, and the at least one reference comprises a reference to a third data item stored in the data storage system.

15. The method as in claim 14, wherein each of the first data item, the second data item, and the third data item are stored on different data storage nodes in the data storage system.

16. The method as in claim 11, wherein collection objects in the data storage system each comprise at least one of a reference to another collection object or a reference to a data file.

17. The method as in claim 16, wherein the collection objects in the data storage system each represent a folder or subfolder of the distributed directory structure.