Memory access communications through message passing interface implemented in memory systems
A memory system having a plurality of memory components and a controller, operatively coupled to the plurality of memory components to: store data in the memory components; communicate with a host system via a bus; service the data to the host system via communications over the bus; communicate with a processing device that is separate from the host system using a message passing interface over the bus; and provide data access to the processing device through communications made using the message passing interface over the bus.
1. A memory system, comprising:
a plurality of memory components; and
a controller, operatively coupled to the plurality of memory components, to at least:
store data in the memory components;
communicate with a host system via a bus;
service the data to the host system via communications over the bus;
communicate the data with a processing device that is separate from the host system using a message passing interface over the bus;
predict a data movement over the bus using a machine learning model, the machine learning model trained using data usage information corresponding to historical data access requests transferring data between the memory components; and
provide data access to the processing device through communications made using the message passing interface over the bus, the data access provided based on the predicted data movement and performed without involving the host system.
2. The memory system of claim 1 , wherein the controller communicates with the processing device without involving the host system.
3. The memory system of claim 2 , wherein the host system includes a central processing unit (CPU).
4. The memory system of claim 3 , wherein the bus is a peripheral component interconnect express (PCIe) bus.
5. The memory system of claim 3 , wherein the processing device is a graphics processing unit (GPU).
6. The memory system of claim 3 , wherein the processing device is a controller of a storage device coupled to the bus.
7. The memory system of claim 3 , wherein the processing device is a controller of a storage device coupled to the bus through a computer network.
8. The memory system of claim 7 , wherein the bus is connected to the computer network via an InfiniBand (IB) interface.
9. The memory system of claim 8 , wherein the memory system is a solid state drive.
10. The memory system of claim 9 , wherein the plurality of memory components includes dynamic random access memory.
11. The memory system of claim 10 , wherein the controller performs predictive data movements to cache, in the dynamic random access memory, a portion of data stored on the storage device.
12. The memory system of claim 11 , wherein the predictive data movements are based on a logical volume configured in a virtual machine running on the host system.
13. The memory system of claim 12 , wherein a portion of the logical volume is in the storage device; and a portion of the logical volume is in the plurality of memory components.
14. The memory system of claim 12 , wherein the controller comprises a data orchestrator implemented using a field programmable gate array (FPGA) or an application specific integrated circuit (ASIC) to implement an artificial neural network.
15. The memory system of claim 12 , wherein the data orchestrator predicts the data movements using the artificial neural network.
16. The memory system of claim 12 , wherein the data orchestrator trains the artificial neural network.
17. A method implemented in a memory system having a plurality of memory components and a controller, the method comprising:
storing data in the memory components;
communicating with a host system via a bus;
servicing the data to the host system via the communicating;
communicating the data with a processing device that is separate from the host system using a message passing interface over the bus;
predicting a data movement over the bus using a machine learning model, the machine learning model trained using data usage information corresponding to historical data access requests transferring data between the memory components; and
providing data access to the processing device through communications made using the message passing interface over the bus, the data access provided based on the predicted data movement and performed without involving the host system.
18. A non-transitory computer storage medium storing instructions which, when executed by a memory system having a plurality of memory components and a controller, cause the memory system to perform a method, the method comprising:
storing data in the memory components;
communicating with a host system via a peripheral bus;
servicing the data to the host system via the communicating;
communicating the data with a processing device that is separate from the host system over the peripheral bus;
predicting a data movement over the bus using a machine learning model, the machine learning model trained using data usage information corresponding to historical data access requests transferring data between the memory components; and
providing data access to the processing device through communications made over the peripheral bus, the data access provided based on the predicted data movement and performed without involving the host system.