Efficient data deployment for a parallel data processing system

1. A method for deploying a data block comprising:

at a virtualization platform running a parallel processing application that includes one or more virtual data nodes:

receiving a first command to write a data block to a storage device;

determining whether the first command was sent by a first virtual data node; and

if the first command was sent by a first virtual data node:

writing the data block to a first location in the storage device,

returning the first location to the first virtual data node,

determining whether the data should be replicated, and

if the data should be replicated, instructing the storage device to internally make a copy of the data block to a second location in the storage device and storing the second location in a tracking structure.

2. The method of claim 1 wherein determining whether the first command was sent by a first virtual data node comprises inspecting the storage command for an indication that the data was sent by a virtual data node, and determining whether the data should be replicated comprises inspecting the storage command for an indication that the data should be replicated.

3. The method of claim 1 further comprising:

receiving a second command to write the data block to a storage device;

determining that the second command was sent by a second virtual data node; and

if the first command was sent by a first virtual data node:

determining that the data block was already replicated on the storage device, and

if the data block was already replicated, returning the location of the copy of the data block stored in the tracking structure to the second virtual data node without writing the data block to the storage device.

4. The method of claim 3 wherein determining whether the second command was sent by a second virtual data node comprises inspecting the storage command for an indication that the data was sent by a virtual data node, and determining whether the data should be replicated comprises inspecting the tracing structure for an entry that maps the second virtual data node to the location of the copy of the data block.

5. The method of claim 2 further comprising:

receiving a third command to read the data block from the storage device;

determining whether the number of pending I/O's requests for the data block exceeds a threshold; and

if the number of pending I/O requests exceeds some threshold:

determining that a copy of the data block exists where the number of pending I/O's for the data block is below the threshold, and

returning the location of the copy of the data block stored in the tracking structure.

6. The method of claim 5 wherein determining the number of pending I/O's for the data block comprises measuring the size of the I/O queue depth of storage device.

7. A computer system for deploying a data block comprising:

a processor;

a volatile memory;

a nonvolatile storage device; and

a non-transitory computer readable storage medium having stored thereon program code that, when executed by the processor, causes the processor to:

at a virtualization platform running a parallel processing application that includes one or more virtual data nodes:

receiving a first command to write a data block to a storage device;

determining whether the first command was sent by a first virtual data node; and

if the first command was sent by a first virtual data node:

writing the data block to a first location in the storage device,

returning the first location to the first virtual data node,

determining whether the data should be replicated, and

if the data should be replicated, instructing the storage device to internally make a copy of the data block to a second location in the storage device and storing the second location in a tracking structure.

8. The computer system of claim 7 wherein determining whether the first command was sent by a first virtual data node comprises inspecting the storage command for an indication that the data was sent by a virtual data node, and determining whether the data should be replicated, comprises inspecting the storage command for an indication that the data should be replicated.

9. The computer system of claim 7 further comprising:

receiving a second command to write the data block to a storage device;

determining that the second command was sent by a second virtual data node; and

if the first command was sent by a first virtual data node:

determining that the data block was already replicated on the storage device, and

if the data block was already replicated, returning the location of the copy of the data block stored in the tracking structure to the second virtual data node without writing the data block to the storage device.

10. The computer system of claim 9 wherein determining whether the second command was sent by a second virtual data node comprises inspecting the storage command for an indication that the data was sent by a virtual data node, and determining whether the data should be replicated comprises inspecting the tracing structure for an entry that maps the second virtual data node to the location of the copy of the data block.

11. The computer system of claim 8 further comprising:

receiving a third command to read the data block from the storage device;

determining whether the number of pending POs requests for the data block exceeds a threshold; and

if the number of pending I/O requests exceeds some threshold:

determining that a copy of the data block exists where the number of pending I/O's for the data block is below the threshold, and

returning the location of the copy of the data block stored in the tracking structure.

12. The computer system of claim 11 wherein determining the number of pending I/O's for the data block comprises measuring the size of the I/O queue depth of storage device.

13. A non-transitory computer readable storage medium having stored thereon program code executable by computer system, the program code embodying a method for deploying a data block comprising:

at a virtualization platform running a parallel processing application that includes one or more virtual data nodes:

receiving a first command to write a data block to a storage device;

determining whether the first command was sent by a first virtual data node; and

if the first command was sent by a first virtual data node:

writing the data block to a first location in the storage device,

returning the first location to the first virtual data node,

determining whether the data should be replicated, and

if the data should be replicated, instructing the storage device to internally make a copy of the data block to a second location in the storage device and storing the second location in a tracking structure.

14. The non-transitory computer readable storage medium of claim 13 wherein determining whether the first command was sent by a first virtual data node comprises inspecting the storage command for an indication that the data was sent by a virtual data node, and determining whether the data should be replicated comprises inspecting the storage command for an indication that the data should be replicated.

15. The non-transitory computer readable storage medium of claim 13 further comprising:

receiving a second command to write the data block to a storage device;

determining that the second command was sent by a second virtual data node; and

if the first command was sent by a first virtual data node:

determining that the data block was already replicated on the storage device, and

if the data block was already replicated, returning the location of the copy of the data block stored in the tracking structure to the second virtual data node without writing the data block to the storage device.

16. The non-transitory computer readable storage medium of claim 15 wherein determining whether the second command was sent by a second virtual data node comprises inspecting the storage command for an indication that the data was sent by a virtual data node, and determining whether the data should be replicated comprises inspecting the tracing structure for an entry that maps the second virtual data node to the location of the copy of the data block.

17. The non-transitory computer readable storage medium of claim 14 further comprising:

receiving a third command to read the data block from the storage device;

determining whether the number of pending I/O's requests for the data block exceeds a threshold; and

if the number of pending I/O requests exceeds some threshold:

determining that a copy of the data block exists where the number of pending I/O's for the data block is below the threshold, and

returning the location of the copy of the data block stored in the tracking structure.

18. The non-transitory computer readable storage medium of claim 17 wherein determining the number of pending I/O's for the data block comprises measuring the size of the I/O queue depth of storage device.