Techniques for implementing remote direct memory access through a data processing unit

Patent metadata

Field	Value
Publication number	US-2025156077-A1
Application number	US-202418943133-A
Country	US
Kind code	A1
Filing date	Nov 11, 2024
Priority date	Nov 11, 2023
Publication date	May 15, 2025
Grant date	—

Abstract

In various embodiments, a proxy application receives a first storage request from a proxy driver executing on a host node. The first storage request indicates a location within the shared storage system and a first address range associated with the host node. The proxy application converts the first storage request to a second storage request that indicates the location and a second address range associated with a proxy node. The proxy application transmits the second storage request to a storage driver that executes on the proxy node and is associated with the shared storage system. The storage driver invokes a remote direct memory access data transfer operation between the shared storage system and the host node to fulfill the first storage request.

First claim

What is claimed is:

1 . A computer-implemented method for processing requests to access a shared storage system, the method comprising:
receiving a first storage request from a proxy driver executing on a host node, wherein the first storage request indicates a location within the shared storage system and a first address range associated with the host node; converting the first storage request to a second storage request that indicates the location and a second address range associated with a proxy node; and transmitting the second storage request to a storage driver that executes on the proxy node and is associated with the shared storage system, wherein the storage driver invokes a remote direct memory access (RDMA) data transfer operation between the shared storage system and the host node to fulfill the first storage request.

2 . The computer-implemented method of claim 1 , wherein converting the first storage request to the second storage request comprises mapping the first address range to the second address range.

3 . The computer-implemented method of claim 1 , wherein converting the first storage request to the second storage request comprises generating a shadow buffer that resides on the proxy node and represents a host buffer that resides on the host node and corresponds to the first address range.

4 . The computer-implemented method of claim 1 , wherein transmitting the second storage request to the storage driver comprises causing a virtual file system executing on the proxy node to route the second storage request to the storage driver based on the location.

5 . The computer-implemented method of claim 1 , wherein invoking the RDMA data transfer operation comprises converting the second storage request to an RDMA request that indicates the location, the first address range, and a remote key that exposes at least a portion of a physical memory included in the host node to the shared storage system.

6 . The computer-implemented method of claim 5 , wherein the remote key comprises a cross guest virtual machine identifier memory key that maps a host address space from a first protection domain associated with the host node to a second protection domain associated with the proxy node.

7 . The computer-implemented method of claim 1 , wherein invoking the RDMA data transfer operation comprises mapping the second address range to the first address range and a remote key that exposes at least a portion of a physical memory included in the host node to the shared storage system.

8 . The computer-implemented method of claim 1 , wherein invoking the RDMA data transfer operation comprises causing data to be copied from the shared storage system to the host node in accordance with the first storage request.

9 . The computer-implemented method of claim 1 , wherein the first storage request is issued by a software application executing on either a central processing unit included in the host node or a graphics processing unit included in the host node.

10 . The computer-implemented method of claim 1 , wherein the shared storage system comprises at least one of shared file storage, shared block storage, or object storage.

11 . One or more non-transitory computer readable media including instructions that, when executed by one or more processors, cause the one or more processors to process requests to access a shared storage system by performing the steps of:
receiving a first storage request from a proxy driver executing on a host node, wherein the first storage request indicates a location within the shared storage system and a first address range associated with the host node; converting the first storage request to a second storage request that indicates the location and a second address range associated with a data processing unit (DPU); and transmitting the second storage request to a storage driver that executes on the DPU and is associated with the shared storage system, wherein the storage driver invokes a remote direct memory access (RDMA) data transfer operation between the shared storage system and the host node to fulfill the first storage request.

12 . The one or more non-transitory computer readable media of claim 11 , wherein converting the first storage request to the second storage request comprises mapping the first address range to the second address range.

13 . The one or more non-transitory computer readable media of claim 11 , wherein converting the first storage request to the second storage request comprises generating a shadow buffer that resides on the DPU and represents a host buffer that resides on the host node and corresponds to the first address range.

14 . The one or more non-transitory computer readable media of claim 11 , wherein the second storage request includes at least one of at least one of a path name or a file handle that corresponds to the location.

15 . The one or more non-transitory computer readable media of claim 11 , wherein invoking the RDMA data transfer operation comprises converting the second storage request to an RDMA request that indicates the location, the first address range, and a remote key that exposes at least a portion of a physical memory included in the host node to the shared storage system.

16 . The one or more non-transitory computer readable media of claim 15 , wherein the remote key indicates the first address range and is subordinate to a cross guest virtual machine identifier memory key.

17 . The one or more non-transitory computer readable media of claim 11 , wherein invoking the RDMA data transfer operation comprises mapping the second address range to the first address range and a remote key that exposes at least a portion of a physical memory included in the host node to the shared storage system.

18 . The one or more non-transitory computer readable media of claim 11 , wherein invoking the RDMA data transfer operation comprises causing data to be copied from the host node to the shared storage system in accordance with the first storage request.

19 . The one or more non-transitory computer readable media of claim 11 , wherein the second storage request is issued by a proxy application that resides in a user space of the DPU.

20 . A system comprising:
one or more memories storing instructions; and one or more processors coupled to the one or more memories that, when executing the instructions, perform the steps of:
receiving a first storage request from a proxy driver executing on a host node, wherein the first storage request indicates a location within a shared storage system and a first address range associated with the host node;
converting the first storage request to a second storage request that indicates the location and a second address range associated with a data processing unit (DPU); and
transmitting the second storage request to a storage driver that executes on the DPU and is associated with the shared storage system, wherein the storage driver invokes a remote direct memory access data transfer operation between the shared storage system and the host node to fulfill the first storage request.

Assignees

• Nvidia Corp

Inventors

• Newburn Chris J
• Bar-Ilan Eliav
• Duer Oren
• Gurtovoy Max
• Modukuri Kiran Kumar
• Shalom Gal
• Spiegelman Roman
• Venkatesan Vishwanath
• Yehezkel Aviad
• Zack Yorey

Classifications

• G06F12/1475

  in a virtual system, e.g. with translation means · CPC title

• G06F2212/1052

  Security improvement · CPC title

• G06F15/17331

  Distributed shared memory [DSM], e.g. remote direct memory access [RDMA] · CPC title

• G06F3/0611 Primary

  in relation to response time · CPC title

• G06F3/0659

  Command handling arrangements, e.g. command buffers, queues, command scheduling · CPC title