Disaggregated computing for distributed confidential computing environment
An apparatus to facilitate disaggregated computing for a distributed confidential computing environment is disclosed. The apparatus includes one or more processors to: provide a remote GPU middleware layer to act as a proxy for an application stack on a client platform that is separate from the remote server platform, wherein the remote GPU middleware layer comprises is to expose an abstraction of the remote GPU to userspace components of a remote GPU stack, the userspace components running on the client machine; communicate with a kernel mode driver of the one or more processors to cause the host memory to be allocated for data structures used to communicate commands between the client and the remote GPU; and invoke the kernel mode driver to submit a workload generated by the application stack, the workload submitted for processing by the remote GPU using the data structures allocated in the host memory.
1. An apparatus comprising:
one or more processors of a remote server platform communicably coupled to host memory and to a remote graphics processing unit (GPU), the one or more processors to:
provide a remote GPU middleware layer on the remote server platform to act as a proxy for an application stack hosted on a client platform that is separate from the remote server platform, wherein the remote GPU middleware layer is to expose an abstraction of the remote GPU to userspace components of a remote GPU stack, the userspace components hosted on the client platform;
communicate, by the remote GPU middleware layer, with a kernel mode driver of the one or more processors to cause the host memory to be allocated for data structures used to communicate commands between the client platform and the remote GPU; and
invoke, by the remote GPU middleware layer, the kernel mode driver to submit a workload generated by the application stack, the workload submitted for processing by the remote GPU using the data structures allocated in the host memory.
2. The apparatus of claim 1 , wherein the data structures comprise command buffers and other data structures that are received from a runtime component and user mode driver component of the client platform, and wherein the command buffers and the other data structures are generated based on instructions from the application stack.
3. The apparatus of claim 2 , wherein the kernel mode driver utilizes the command buffers and the other data structures to prepare a context of the workload and schedule the workload on the remote GPU.
4. The apparatus of claim 1 , wherein the client platform is to execute a corresponding remote GPU middleware layer to interface with the remote GPU middleware layer of the remote server platform, and wherein the remote GPU middleware layer is to expose the abstraction of the remote GPU to the userspace components of the remote GPU stack on the client platform, and is to mediate transfer of data between the client platform and the remote GPU.
5. The apparatus of claim 1 , wherein the remote GPU middleware layer is a transport-agnostic interface for the application stack on the client platform.
6. The apparatus of claim 1 , wherein the remote GPU middleware layer comprises a transport sublayer to communicate command and data between the client platform and the remote GPU.
7. The apparatus of claim 1 , wherein the remote GPU comprises a network interface controller (NIC) for direct transfers of data between the client platform and the remote GPU.
8. The apparatus of claim 1 , wherein a GPU local memory of the remote GPU is mapped to an address space of the application stack of the client platform to allow the application stack to access the GPU local memory directly.
9. The apparatus of claim 1 , wherein the one or more processors comprise one or more of a GPU, a central processing unit (CPU), or a hardware accelerator.
10. A method comprising:
providing, by one or more processors of a remote server platform communicably coupled to a host memory and to a remote graphics processing unit (GPU), a remote GPU middleware layer on the remote server platform to act as a proxy for an application stack hosted on a client platform that is separate from the remote server platform, wherein the remote GPU middleware layer is to expose an abstraction of the remote GPU to userspace components of a remote GPU stack, the userspace components hosted on the client platform;
communicating, by the remote GPU middleware layer, with a kernel mode driver of the one or more processors to cause the host memory to be allocated for data structures used to communicate commands between the client platform and the remote GPU; and
invoking, by the remote GPU middleware layer, the kernel mode driver to submit a workload generated by the application stack, the workload submitted for processing by the remote GPU using the data structures allocated in the host memory.
11. The method of claim 10 , wherein the data structures comprise command buffers and other data structures that are received from a runtime component and user mode driver component of the client platform, and wherein the command buffers and the other data structures are generated based on instructions from the application stack.
12. The method of claim 11 , wherein the kernel mode driver utilizes the command buffers and the other data structures to prepare a context of the workload and schedule the workload on the remote GPU.
13. The method of claim 10 , wherein the client platform is to execute a corresponding remote GPU middleware layer to interface with the remote GPU middleware layer of the remote server platform, and wherein the remote GPU middleware layer is to expose the abstraction of the remote GPU to the userspace components of the remote GPU stack on the client platform, and is to mediate transfer of data between the client platform and the remote GPU.
14. The method of claim 10 , wherein the remote GPU middleware layer comprises a transport sublayer to communicate command and data between the client platform and the remote GPU.
15. The method of claim 10 , wherein a GPU local memory of the remote GPU is mapped to an address space of the application stack of the client platform to allow the application stack to access the GPU local memory directly.
16. A non-transitory machine readable storage medium having stored thereon executable computer program instructions that, when executed by one or more processors, cause the one or more processors to perform operations to:
provide, by the one or more processors of a remote server platform communicably coupled to a host memory and to a remote graphics processing unit (GPU), a remote GPU middleware layer on the remote server platform to act as a proxy for an application stack hosted on a client platform that is separate from the remote server platform, wherein the remote GPU middleware layer is to expose an abstraction of the remote GPU to userspace components of a remote GPU stack, the userspace components hosted on the client platform;
communicate, by the remote GPU middleware layer, with a kernel mode driver of the one or more processors to cause the host memory to be allocated for data structures used to communicate commands between the client platform and the remote GPU; and
invoke, by the remote GPU middleware layer, the kernel mode driver to submit a workload generated by the application stack, the workload submitted for processing by the remote GPU using the data structures allocated in the host memory.
17. The non-transitory machine readable storage medium of claim 16 , wherein the data structures comprise command buffers and other data structures that are received from a runtime component and user mode driver component of the client platform, and wherein the command buffers and the other data structures are generated based on instructions from the application stack.
18. The non-transitory machine readable storage medium of claim 17 , wherein the kernel mode driver utilizes the command buffers and the other data structures to prepare a context of the workload and schedule the workload on the remote GPU.
19. The non-transitory machine readable storage medium of claim 16 , wherein the client platform is to execute a corresponding remote GPU middleware layer to interface with the remote GPU middleware layer of the remote server platform, and wherein the remote GPU middleware layer is to expose the abstraction of the remote GPU to the userspace components of the remote GPU stack on the client platform, and is to mediate transfer of data between the client platform and the remote GPU.
20. The non-transitory machine readable storage medium of claim 16 , wherein a GPU local memory of the remote GPU is mapped to an address space of the application stack of the client platform to allow the application stack to access the GPU local memory directly.