Shared Memory Fabric Orchestration Using DAG Resource Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional workload performance systems inefficiently utilize processing, memory, and storage resources due to sequential data retrieval and storage across multiple server devices, leading to suboptimal resource utilization.
Innovation Solution
A shared memory fabric workload performance system utilizing a resource orchestrator engine that generates a Directed Acyclic Graph (DAG) to map functions to processing and memory resources, optimizing data transfers and resource usage through a distributed tiered network fabric.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional sequential workload performance operations are used across multiple server devices, then workload operations can be performed, but resource utilization becomes inefficient
Solution Approach 1:
The patent segments the workload into multiple independent functions that can be executed in parallel across different processing resources. Each function is identified and mapped to specific processing and memory resources, allowing simultaneous execution rather than sequential processing. This segmentation enables multiple server devices to work concurrently on different functions of the same workload, dramatically improving resource utilization efficiency.
2Productivity
If data is retrieved and stored sequentially across multiple server devices, then workload operations can complete, but data transfer efficiency decreases
Solution Approach 1:
The patent implements preliminary action by pre-mapping functions to processing and memory resources before execution. The system identifies all required functions, determines their resource requirements, and establishes the execution plan in advance. This allows data to be positioned in optimal memory locations beforehand and processing resources to be pre-configured, eliminating unnecessary data movement during execution and reducing overall data transfer time.
3Productivity
If processing resources are distributed across multiple server devices, then workload capacity increases, but resource coordination complexity increases
Solution Approach 1:
The patent introduces an intermediary component that acts as a central coordinator between distributed processing resources. This intermediary manages the mapping of functions to processing and memory resources, handles resource allocation, and coordinates data flow between different server devices. By centralizing the coordination logic in this intermediary layer, the system can manage complex multi-device workloads without requiring each device to independently coordinate with every other device, thus reducing overall system complexity.
Data Source
AI summary
A shared memory fabric workload performance system includes a resource orchestrator device coupled to processing systems and memory systems that are configured to provide a shared memory fabric to each of the processing systems. The resource orchestrator device receives a request to perform a workload, identifies functions for performing the workload, and generates a DAG that identifies a respective processing resource type and a respective memory requirement for performing each of the functions. For each of the functions, the resource orchestrator device determines a respective processing resource provided by the processing systems that includes the processing resource type identified in the DAG for performing that function, and a respective memory resource provided by the memory systems that is accessible to that respective processing resource, and maps that respective memory resource to that respective processing resource based on the memory requirement identified in the DAG for performing that function.