公开(公告)号：US20250133491A1

公开(公告)日：2025-04-24

申请号：US18381367

申请日：2023-10-18

Applicant: Google LLC

Inventor： Ananya Simlai ,  Rittwik Jana ,  Ian Kenneth Coolidge ,  Santanu Dasgupta

IPC: H04W52/02

Abstract: Aspects of the disclosure are directed to network optimization of various workload servers running in a distributed cloud platform through closed loop machine learning inferencing performed locally on the workload servers. The workload servers can each be equipped with one or more machine learning accelerators to respectively perform local predictions for the workload servers. In response to the local predictions, attributes of the workload servers can be adjusted automatically for optimizing the network.

公开(公告)号：US20250123673A1

公开(公告)日：2025-04-17

申请号：US18379798

申请日：2023-10-13

Applicant: Google LLC

Inventor： Ananya Simlai ,  Ming Wen ,  Ian Kenneth Coolidge ,  Santanu Dasgupta

IPC: G06F1/329

Abstract: The presently disclosed technology provides methods and systems for optimally allocating power among workloads executing on a computer system through use of a power management algorithm. For example, according to the present technology a plurality of CPUs within a server can be divided into multiple groups according to application workloads. Workloads can be distributed to the CPUs as needed by a workload scheduler, and the workload scheduler can provide the CPU IDs to a power manager, enabling the power manager to optimize power settings. Each group of CPUs can be assigned an optimal power profile tailored to its respective situation.

公开(公告)号：US20230418775A1

公开(公告)日：2023-12-28

申请号：US18213028

申请日：2023-06-22

Applicant: Google LLC

Inventor： Santanu Dasgupta ,  Bok Knun Randolph Chung ,  Ankur Jain ,  Prashant Chandra ,  Bor Chan ,  Durgaprasad V. Ayyadevara ,  Ian Kenneth Coolidge ,  Muzammil Mueen Butt

IPC: G06F13/38 ,  G06F13/28

CPC classification number: G06F13/385 ,  G06F13/28 ,  G06F2213/0038 ,  G06F2213/3808 ,  G06F2213/0026

Abstract: The present disclosure provides for a converged compute platform architecture, including a first infrastructure processing unit (IPU)-only configuration and a second configuration wherein the IPU is coupled to a central processing unit, such as an x86 processor. Connectivity between the two configurations may be accomplished with a PCIe switch, or the two configurations may communicate through remote direct memory access (RDMA) techniques. Both configurations may use ML acceleration through a single converged architecture.

公开(公告)号：US20230267089A1

公开(公告)日：2023-08-24

申请号：US18112740

申请日：2023-02-22

Applicant: Google LLC

Inventor： Santanu Dasgupta ,  Durgaprasad V. Ayyadevara ,  Bor Chan ,  Prashant R. Chandra ,  Bok Knun Randolph Chung ,  Max Kamenetsky ,  Rajeev Koodli ,  Shahin Valoth

IPC: G06F13/38 ,  G06F13/12

CPC classification number: G06F13/385 ,  G06F13/122 ,  G06F2213/0038

Abstract: The present disclosure provides a compute platform architecture for virtualized and cloud native network functions. The architecture uses a reduced instruction set computer-based general purpose processor along with multiple special purpose accelerators and an integrated network interface card. As such, the architecture can accommodate multiple hundreds of gigabits of input/output.