Inter-Kernel Communication Abstraction for Accelerator Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Accelerator device kernels in data centers face limitations in accessing and communicating with a large percentage of other accelerator devices due to restricted communication faculties, hindering efficient workload execution in disaggregated resource environments.
Innovation Solution
The implementation of an inter-kernel communication system using application programming interfaces (APIs) and orchestrator servers that manage and configure accelerator resources to establish logical communication paths across various physical paths, enabling efficient inter-kernel communication and resource allocation based on quality of service parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If accelerator device kernels are configured to establish connections with other accelerator device kernels through physical communication paths, then communication between accelerator devices is enabled, but the communication capability is limited to single types of accelerator devices and a large percentage of accelerator devices in the data center cannot be accessed
Solution Approach 1:
The patent introduces communication abstraction logic units as intermediary components that sit between accelerator device kernels and physical communication paths. These abstraction layers translate diverse physical communication protocols into a unified logical communication interface, enabling accelerator device kernels to communicate with multiple types of accelerator devices without direct configuration of communication protocols. The abstraction logic units manage the complexity of inter-kernel communication while presenting a simplified interface to kernels.
2Productivity
If communication protocols are detailed and configured in the kernel by the kernel developer, then communication with specific accelerator device kernels is established, but the primary purpose of performing operations efficiently is compromised due to limited communication faculties
Solution Approach 1:
The patent extracts communication protocol configuration and management functions from the accelerator device kernels themselves and places them in dedicated communication abstraction logic units. This separation allows kernels to focus exclusively on performing computational operations efficiently, while communication responsibilities are handled by the extracted abstraction layer. The kernels no longer need to contain detailed communication protocol configurations, simplifying kernel development and improving operational efficiency.
3Adaptability or versatility
If accelerator devices utilize pools of disaggregated resources through various physical communication paths, then resource utilization is improved, but a given accelerator device kernel remains unable to access a large percentage of other accelerator devices
Solution Approach 1:
The patent implements universal communication abstraction logic units that can handle multiple types of physical communication paths and protocols through a single unified interface. This multi-functional abstraction layer enables any accelerator device kernel to access any other accelerator device in the data center regardless of the underlying physical communication path or protocol. The universal abstraction logic provides a single interface that adapts to diverse communication requirements, dramatically increasing the quantity of accessible accelerator devices from each kernel.
Data Source
AI summary
Technologies for providing inter-kernel communication application programming interfaces (API) include an orchestrator device comprising circuitry to receive a request to allocate one or more accelerator resources to a given workload. The circuitry is also configured to identify one or more kernel bit streams in the accelerator resources used to perform the workload. The circuitry is configured to determine, from the identified one or more kernel bit streams, an inter-kernel communication topology and configure the identified one or more kernel bit streams according to the inter-kernel communication topology.


