Memory Appliance Bandwidth Allocation for Legacy Processors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-processor computer systems with PCIe interfaces are limited in exploiting the advantages of memory fabrics, such as increased memory access and lower latency, due to inefficiencies in memory bandwidth allocation and the inability to perform collective memory operations quickly.
Innovation Solution
Dynamic allocation of memory bandwidth and offloading of certain operations from legacy processors to a memory appliance with high-speed connections to the memory fabric, allowing for parallel execution of collective memory operations and improved resource management through a scheduler and memory appliance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If PCIe interfaces are used for memory access, then compatibility with legacy processors is maintained, but memory access speed and bandwidth are limited
Solution Approach 1:
A memory appliance is introduced as an intermediary component between legacy processors and the memory fabric. The memory appliance includes a PCIe interface that communicates with legacy processors and a high-speed interface that connects to the memory fabric, enabling fast memory access while maintaining legacy processor compatibility.
Solution Approach 2:
The system is segmented into distinct components: legacy processors, memory appliance, and memory fabric. The memory appliance acts as a bridge that separates the legacy PCIe interface requirements from the high-speed memory fabric requirements, allowing each component to operate at its optimal performance level.
2Productivity
If static memory bandwidth allocation is used, then resource allocation is simple, but processing efficiency decreases due to processor stalling
Solution Approach 1:
The system implements dynamic memory bandwidth allocation where the memory appliance can adjust bandwidth allocation in real-time based on the actual needs of computing jobs. This dynamic adjustment prevents processor stalling by ensuring adequate bandwidth is available when needed, while maintaining manageable complexity through automated allocation algorithms.
Solution Approach 2:
The system employs feedback mechanisms where the memory appliance monitors memory access patterns and bandwidth utilization, then adjusts allocation accordingly. This feedback loop ensures processing efficiency is maintained while keeping resource allocation complexity manageable through automated control.
3Speed
If collective memory operations are performed through PCIe, then processor coordination is maintained, but operation speed is slow due to interconnect limitations
Solution Approach 1:
The memory appliance serves as a mediator for collective memory operations, receiving operation requests from multiple legacy processors through PCIe interfaces and executing them in parallel through the high-speed memory fabric interface, thereby accelerating collective operations while managing coordination complexity.
Solution Approach 2:
Multiple collective memory operations from different processors are merged and executed in parallel through the memory appliance's high-speed interface to the memory fabric, combining the computational power of multiple processors while achieving speeds limited only by the memory fabric capacity.
Data Source
AI summary
A method for allocating resources includes determining that an initial allocation of memory bandwidth for one or more computing jobs fails a performance metric. The memory bandwidth provides access to a global memory pool for multiple legacy processors across a memory fabric. The method also includes determining a new allocation of memory bandwidth for the computing jobs that meets the performance metric. Additionally, the method includes assigning the new allocation of memory bandwidth to the computing jobs. The method further includes executing the computing jobs using the new allocation of memory bandwidth.