Fabric-Attached Memory Reliability Coordination by Application Need
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed computing systems with fabric attached memory face challenges in managing reliability features to optimize system operation, as existing methods often result in inefficiencies due to mismatched reliability requirements, leading to excessive energy consumption, performance degradation, or insufficient reliability.
Innovation Solution
A system-level approach is implemented to coordinate reliability features across different levels of the distributed computing system, including IC die, memory device, and system levels, using error detection and correction mechanisms like ECC, RAID, and data replication, tailored to specific application needs through a lookup table to achieve the desired reliability levels without overprovisioning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If reliability features are configured at maximum level across the distributed computing system, then system reliability is improved, but energy consumption increases and performance degrades
Solution Approach 1:
The patent applies local quality by configuring reliability features at different levels (IC die level, memory device level, system level) based on specific application requirements rather than uniformly across the entire system. This allows each component to have the appropriate level of reliability protection needed, avoiding excessive energy consumption from over-provisioning reliability features throughout the system.
Solution Approach 2:
The patent implements dynamic configuration of reliability features through a library of pre-configured reliability settings that can be selected and applied based on application-specific requirements. This dynamic approach allows the system to adapt reliability levels to match actual needs, preventing both over-provisioning (which wastes energy) and under-provisioning (which compromises reliability).
2Reliability
If reliability features are configured at maximum level across the distributed computing system, then system reliability is improved, but system performance deteriorates
Solution Approach 1:
The patent applies local quality by configuring reliability features at different levels (IC die level, memory device level, system level) based on specific application requirements rather than uniformly across the entire system. This allows each component to have the appropriate level of reliability protection needed, avoiding performance degradation from unnecessary reliability overhead in components that don't require it.
Solution Approach 2:
The patent implements dynamic configuration of reliability features through a library of pre-configured reliability settings that can be selected and applied based on application-specific requirements. This dynamic approach allows the system to adapt reliability levels to match actual needs, preventing performance deterioration from over-provisioning while maintaining adequate reliability where required.
Data Source
AI summary
A method performed by a distributed computing system is described. The distributed computing system includes one or more host nodes and a main memory connected to the one or more host nodes by a fabric interconnect. The method includes receiving, by an application program interface (API) of a host node, a function call from an application of a host device, the function call including a pointer to a memory object and a level of reliability for operations involving the memory object; and configuring system reliability features included in one or both of the fabric interconnect and memory devices of the main memory according to the level of reliability in the function call.