Storage processing unit arrays and methods of use
View Patent ↗The embodiments disclosed herein include an interconnection network that is configured to provide data communication between storage processing units. The disclosed interconnection network can be particularly effective when the storage processing units are configured to locally perform scientific computations. The disclosed interconnection network can enable localized, high throughput, and low latency data communication between storage processing units without overloading the host system.
1. A solid state device system comprising:
a plurality of storage processing units arranged as a two-dimensional array, wherein each of the plurality of storage processing units comprises a computational unit and a non-volatile memory module; and
an interconnection network comprising a plurality of unit network modules, wherein the interconnection network is configured to provide data communication between the plurality of storage processing units;
wherein the plurality of storage processing units is grouped into a plurality of sub-arrays of storage processing units;
wherein storage processing units in a first sub-array of the plurality of sub-arrays are coupled to one another using a first one of the plurality of unit network modules; and
a memory controller configured to schedule data transfer between the plurality of storage processing units via the interconnection network.
2. The system of claim 1 , wherein storage processing units in a second sub-array of the plurality of sub-arrays are coupled to one another using a second one of the plurality of unit network modules.
3. The system of claim 1 , wherein the memory controller is further configured to:
load a plurality of blocks corresponding to a two-dimensional matrix onto the plurality of sub-arrays;
trigger the plurality of sub-arrays to locally perform a transpose operation to compute a transpose of the plurality of blocks; and
trigger two of the sub-arrays to exchange the transpose of blocks stored in the two of the sub-arrays via the interconnection network.
4. The system of claim 1 , wherein the computational unit in the plurality of storage processing units is configured to perform scientific computation.
5. The system of claim 1 , wherein one of the plurality of unit network modules comprises a L-diagonal type 1 unit network module.
6. The system of claim 1 , wherein one of the plurality of unit network modules comprises a L-diagonal type 2 unit network module.
7. The system of claim 1 , wherein one of the plurality of unit network modules comprises a L-diagonal type 3 unit network module.
8. The system of claim 1 , wherein one of the plurality of unit network modules comprises a L-diagonal type 4 unit network module.
9. The system of claim 1 , wherein one of the plurality of unit network modules comprises a full-mesh unit network module.
10. A system comprising:
the solid state device system of claim 1 ; and
a host device in data communication with the solid state device system, wherein the host device is configured to send data to the solid state device system to be processed by the computational unit in the plurality of storage processing units.
11. The system of claim 2 , wherein the plurality of sub-arrays, including the first sub-array and the second sub-array, are coupled to one another using a third one of the plurality of unit network modules, thereby forming a hierarchical interconnection of storage processing units.
12. The system of claim 3 , wherein the memory controller is configured to determine a route on the interconnection network over which the two of the sub-arrays exchange the transpose of blocks.
13. The system of claim 11 , wherein each of the unit network modules has a first type of interconnection topology.
14. The system of claim 11 , wherein the first one of the plurality of unit network modules has a first type of interconnection topology and the second one of the plurality of unit network modules has a second type of interconnection topology.
15. The system of claim 11 , wherein the first sub-array is disposed diagonally from the second sub-array, and the third one of the plurality of unit network modules comprises a diagonal interconnect that is configured to directly connect the first sub-array and the second sub-array.
16. The system of claim 13 , wherein the first type of interconnection topology comprises a modified ring network topology.
17. The system of claim 14 , wherein the first type of interconnection topology comprises a modified ring network topology.
18. The system of claim 15 , wherein the diagonal interconnect is configured to directly connect a communication hub of the first sub-array and a communication hub of the second sub-array.
19. The system of claim 1 , wherein the plurality of storage processing units is arranged on printed circuit board and the interconnection network comprises an electrical conductor on the printed circuit board.