Systems and methods for synchronous cell switching for scalable memory
A system includes: a group of memory resources including a first memory node and a second memory node, the first memory node being connected to the second memory node over a switching fabric; and a synchronous clock source connected to the first memory node and the second memory node, the synchronous clock source to provide a synchronized clock signal to the first memory node and the second memory node to synchronize the first memory node with the second memory node. The first memory node and the second memory node are to encode memory data and decode encoded memory data using the synchronized clock signal.
1. A system comprising:
a group of memory resources comprising a first memory node and a second memory node, the first memory node being connected to the second memory node over a switching fabric; and
a synchronous clock source connected to the first memory node and the second memory node, the synchronous clock source being configured to generate a synchronized clock signal based on a reference clock signal, and provide the synchronized clock signal to the first memory node and the second memory node to synchronize the first memory node with the second memory node,
wherein the first memory node and the second memory node are configured to encode memory data and decode encoded memory data using the synchronized clock signal received from the synchronized clock source.
2. The system of claim 1 , wherein:
the first memory node comprises a first controller, the first controller being configured to encode the memory data using the synchronized clock signal, and transmit the encoded memory data to the second memory node; and
the second memory node comprises a second controller, the second controller being configured to receive the encoded memory data from the first memory node, and decode the encoded memory data using the synchronized clock signal.
3. The system of claim 1 , further comprising a memory group controller connected to:
the first memory node and the second memory node via the switching fabric, the switching fabric comprising a cache coherent protocol-based interconnect; and
an input/output (I/O) connection,
wherein the memory group controller is configured to allocate memory from at least one of the first memory node or the second memory node by enabling communications between the I/O connection and the at least one of the first memory node or the second memory node.
4. The system of claim 1 , further comprising:
a first processor connected to the first memory node and the second memory node of the group of memory resources via the switching fabric, and connected to a first network interface card (NIC); and
a second processor connected to a third memory node and a fourth memory node of the group of memory resources via a second switching fabric, and connected to a second NIC,
wherein the first processor is configured to communicate with the second processor through the first NIC or the second NIC.
5. The system of claim 4 , wherein the first NIC is connected to the second NIC via a serial interface.
6. The system of claim 4 , wherein the third memory node and the fourth memory node are configured to encode memory data and decode encoded memory data using the synchronized clock signal.
7. The system of claim 4 , wherein the first processor comprises one of a system on chip (SoC), a central processing unit (CPU), a graphics processing unit (GPU), a tensor processing unit (TPU), a field-programmable gate array (FPGA), or an application-specific integrated circuit (ASIC).
8. The system of claim 1 , wherein the first memory node comprises a first one of a memory die, a memory module, a memory sled, a memory pod, or a memory rack, and the second memory node comprises a second one of a memory die, a memory module, a memory sled, a memory pod, or a memory rack.
9. The system of claim 8 , wherein the first memory node comprises a first memory rack and the second memory node comprises a second memory rack, the second memory rack being different from the first memory rack.
10. The system of claim 9 , wherein a maximum latency between the first memory node and the second memory node is less than a predetermined threshold.
11. The system of claim 10 , further comprising a processor in communication with a processor memory, the processor memory storing instructions that, when executed by the processor, cause the processor to run an application that stores information in the group of memory resources.
12. The system of claim 1 , wherein the first memory node is further connected to the second memory node over a second switching fabric.
13. The system of claim 1 , wherein the group of memory resources is cache coherent.
14. A system comprising:
a group of memory resources comprising:
a first memory node comprising a first memory controller, a first memory, and a first cache coherent protocol-based interconnect; and
a second memory node comprising a second memory controller, a second memory, and a second cache coherent protocol-based interconnect, the second memory node being connected to the first memory node over a switching fabric connecting the first cache coherent protocol-based interconnect with the second cache coherent protocol-based interconnect; and
a synchronous clock source connected to the first memory node and the second memory node, the synchronous clock source being configured to generate a synchronized clock signal based on a reference clock signal, and provide the synchronized clock signal to the first memory node at the first cache coherent protocol-based interconnect and to the second memory node at the second cache coherent protocol-based interconnect to synchronize the first memory node with the second memory node,
wherein the first memory node and the second memory node are configured to send and receive memory data to and from each other by encoding and decoding the memory data using the synchronized clock signal from the synchronous clock source.
15. The system of claim 14 , wherein the first memory node and the second memory node are cache coherent.
16. The system of claim 14 , wherein the first memory node and the second memory node are configured to encode and decode the memory data that is sent and received from each other using the synchronized clock signal.
17. The system of claim 14 , wherein the synchronous clock source comprises:
a clock synthesizer comprising a phase-locked loop; and
a clock buffer.
18. The system of claim 17 , wherein the clock buffer is connected to the first memory node at the first cache coherent protocol-based interconnect, and is connected to the second memory node at the second cache coherent protocol-based interconnect.
19. A method of encoding and decoding data in separate memory nodes of a group of memory resources, comprising:
storing memory data in a first memory node;
encoding the memory data using a synchronized clock signal at the first memory node, the synchronized clock signal being generated based on a reference clock signal;
sending the encoded memory data to a second memory node via a switching fabric;
receiving the encoded memory data at the second memory node via the switching fabric; and
decoding the encoded memory data using the synchronized clock signal at the second memory node, the synchronized clock signal at the second memory node being the synchronous clock signal used to encode the memory data at the first memory node.
20. The method of claim 19 , wherein:
the switching fabric comprises a cache coherent protocol-based interconnect;
the first memory node comprises a first memory rack comprising a first plurality of memory devices; and
the second memory node comprises a second memory rack comprising a second plurality of memory devices.