Cloud server memory expansion device

Abstract

A cloud server memory expansion device includes: at least one cloud server, comprising at least one central processing unit and a graphics processing unit, at least one random access memory (RAM) with data pages, and at least one storage device; a memory proxy module for monitoring the RAM and managing migration markers of the data pages; a memory control module signal-connected to the memory proxy module, for determining the migration order of the data pages based on the RAM usage and migration strategy; a memory resource pool serving as a remote memory expansion device and connected to the memory control module, comprising a cache storage layer and at least one remote memory node; and a network transmission module connecting the cloud server and the memory resource pool for transmitting the data pages.

Claims

1. A cloud server memory expansion device, characterized in that it comprises: at least one cloud server, which is a cloud virtual machine (VM), including at least one central processing unit (CPU) and graphics processing unit (GPU), at least one random-access memory (RAM) with memory pages, and at least one storage device; a memory agent module, disposed within the cloud server, for monitoring the RAM usage status of the cloud server, intercepting memory shortage signals, and managing the migration flags of the memory pages; a memory controller module, signal-connected to the memory agent module, the memory controller module determining the migration order of the memory pages based on the RAM usage rate and a migration strategy; the memory controller module is configured to be inserted into a Peripheral Component Interconnect (PCI). A circuit board card with one of the following slots: PCIe (Express, PCIe) or Compute Express Link (CXL); electrically connected to the central processing unit, graphics processing unit, and random access memory of the cloud server via a bus, wherein the bus supports the PCIe or CXL protocol; a memory pool, which serves as a remote memory expansion device and is connected to the memory control module, the memory pool including a cache layer and at least one remote memory node for storing the data pages migrated from the cloud server; the cache layer serves as a high-speed buffer when the random access memory is full, and the remote memory node provides the main expanded memory capacity for storing the data pages with lower access frequencies; a network transmission module, connecting the cloud server and the memory pool, for transmitting the data pages; Furthermore, when the utilization rate of the random access memory reaches a preset threshold, the memory control module sequentially migrates the marked data page to the cache storage layer and the remote memory node of the memory resource pool via the network transmission module; wherein, Data pages are moved to the swap space of the storage device only when the memory resource pool is full (i.e., when both the cache storage layer and the remote memory node have reached their capacity limits).

2. The cloud server memory expansion device as described in claim 1, wherein, The cloud server is used on public cloud platforms or private cloud platforms.

3. The cloud server memory expansion device as described in claim 1, wherein, The memory control module uses a migration strategy selected from a Least Recently Used (LRU) algorithm, a Least Frequently Used (LFU) algorithm, or a combination thereof.

4. The cloud server memory expansion device as described in claim 1, wherein, The memory control module is a hardware control circuit, which can be selected from a programmable logic gate (FPGA) or an application-specific integrated circuit (ASIC).

5. The cloud server memory expansion device as described in claim 1, wherein, The memory control module includes a bus controller, an internal buffer memory, and an arithmetic logic unit.

6. The cloud server memory expansion device as described in claim 1, wherein, The remote memory node of the memory pool is selected from a high-memory server or a distributed memory system.

7. The cloud server memory expansion device as described in claim 1, wherein, The network transmission module supports Transmission Control Protocol (TCP) or Remote Direct Memory Access (RDMA) protocol.

Images