DATA STORAGE ARCHITECTURE AND SYSTEM FOR HIGH PERFORMANCE COMPUTING

• US 20140108707A1
• Filed: 09/16/2013
• Published: 04/17/2014
• Est. Priority Date: 10/17/2012
• Status: Active Grant

• Current Assignee: DataDirect Networks Incorporated
• Original Assignee: DataDirect Networks Incorporated
• Inventors: Nowoczynski, Paul, Cope, Jason Micah, Manning, Gordon, Molaro, Don, Piszczek, Michael, Uppu, Pavan
• Granted Patent: US 9,128,826 B2
• Time in Patent Office: Days
• US Class Current: 711/103
• CPC Class Codes: G06F 12/0246 in block erasable memory, e.g. flash memory
• CPC Class Codes: G06F 12/0871 Allocation or management of cache space
• CPC Class Codes: G06F 13/4282 on a serial bus, e.g. I2C bus, SPI bus on daisy chain buses G06F13/4247
• CPC Class Codes: G06F 16/10 File systems; File servers
• CPC Class Codes: G06F 16/137 Hash-based content-based indexing of textual data G06F16/31
• CPC Class Codes: G06F 16/182 Distributed file systems
• CPC Class Codes: G06F 3/061 Improving I/O performance
• CPC Class Codes: G06F 3/0655 Vertical data movement, i.e. input-output transfer; data movement between one or more hosts and one or more storage devices
• CPC Class Codes: G06F 3/0688 Non-volatile semiconductor memory arrays

Abstract

Data storage systems and methods for storing data are described herein. The storage system may be integrated with or coupled with a compute cluster or super computer having multiple computing nodes. A plurality of nonvolatile memory units may be included with computing nodes, coupled with computing nodes or coupled with input/output nodes. The input/output nodes may be included with the compute cluster or super computer, or coupled thereto. The nonvolatile memory units store data items provided by the computing nodes, and the input/output nodes maintain where the data items are stored in the nonvolatile memory units via a hash table distributed among the input/output nodes. The use of a distributed hash table allows for quick access to data items stored in the nonvolatile memory units even as the computing nodes are writing large amounts of data to the storage system quickly in bursts.

Claims

1. A data storage method comprising:

an I/O node receiving a storage request from a computing node of a super computer;

the I/O node computing a hash on metadata for a data item referenced in the storage request to obtain a nonvolatile memory (NVM) location for the data item;

the I/O node creating an entry for the data item with the NVM location in its portion of a distributed hash table;

the I/O node determining if the data item was stored in an expected storage location based on the hash of the metadata of the data item;

if the data item was not stored in the expected storage location,

the I/O node communicating the NVM storage location of the data item to the expected I/O node,

the expected I/O node creating an entry for the data item in its portion of the distributed hash table,

the expected I/O node initiating sending the data item to primary storage,

the I/O node removing the entry for the data item from its portion of the distributed hash table, and

the expected I/O node updating its entry for the data item in its portion of the distributed hash table signifying that the data item has been moved to primary storage;

if the data item was stored in the expected storage location,

the expected I/O node initiating sending the data item to primary storage,

the I/O node updating its portion of the distributed hash table signifying that the data item is available in primary storage.

2. The method of claim 1 wherein the metadata includes a data item identifier and an offset for the data item.

3. The method of claim 1 wherein each entry in the distributed hash table includes either an address of one of the NVM units where the data item is stored or an indication the data item is stored in primary storage.

4. A compute cluster comprising:

a plurality of computing nodes coupled with a high speed interconnect and coupled with a local interconnect;

a plurality of nonvolatile memory units coupled with the local interconnect;

a plurality of input/output nodes coupled with the high speed interconnect, each input/output node including a portion of a distributed hash table to maintain a data item location of data items provided by the computing nodes indexed according to a hash on metadata for the data items.

5. The compute cluster of claim 4 wherein each input/output node is a server computer.

6. The compute cluster of claim 4 wherein a number of computing nodes exceeds a number of nonvolatile memory units.

7. The compute cluster of claim 4 wherein a number of nonvolatile memory units exceeds a number of input/output nodes.

8. The compute cluster of claim 4 wherein the metadata includes a data item identifier and an offset for the data item.

9. The compute cluster of claim 4 wherein each entry in the distributed hash table includes either an address in one of the nonvolatile memory units where the data item is stored or an indication the data item is stored in a primary storage.

10. The compute cluster of claim 4 wherein the compute cluster is a super computer.

11. A data retrieval method comprising:

a computing node of a plurality of computing nodes performing a hash on metadata for a data item;

the computing node sending a read request for the data item to an input/output node (I/O node), the read request including the hash result;

the I/O node receiving the read request for the data item from the computing node;

the I/O node looking for an entry for the data item in its portion of a distributed hash table based on the hash result;

the I/O node obtaining the data item from a non-volatile memory location or from a primary storage;

the I/O node providing the data item to the computing node.

12. The method of claim 11 wherein the I/O node obtaining the data item from the primary storage includes reformatting the read request and forwarding the reformatted read request to the primary storage.

13. The method of claim 11 wherein the I/O node obtaining the data item from the non-volatile memory location includes the I/O node obtaining the data from its own non-volatile memory or group of non-volatile memory under its control and management.

14. The method of claim 11 wherein the I/O node obtaining the data item from the non-volatile memory location includes the I/O node forwarding the read request to another I/O node when the data item resides in the other I/O node'"'"'s non-volatile memory or non-volatile memory under the other I/O node'"'"'s control and management.

15. The method of claim 11 wherein the I/O node obtaining the data item from the non-volatile memory or from the primary storage comprises:

obtaining the data item from a non-volatile memory location included as the entry in the distributed hash table;

obtaining the data item from the primary storage when no entry for the hash of the metadata for the data item in the distributed hash table can be found.

16. The method of claim 11 wherein the entry in the hash table contains either an address in non-volatile memory of the data item or a flag signifying the data item is located in primary storage.

17. The method of claim 11 wherein the computing nodes form a super computer.

18. The method of claim 11 wherein the computing nodes form a compute cluster.