Havens: Explicit Reliable Memory Regions for HPC Applications

TL;DR

Havens introduces a software-based memory protection scheme using fault-protected regions called havens, aimed at enhancing error resilience in future exascale supercomputers with high memory fault rates.

Contribution

The paper proposes a novel region-based memory management approach with havens that provide fault protection for critical program objects in HPC applications.

Findings

Provides fault protection for program objects using havens.

Application-agnostic fault coverage.

Enables placement of critical objects in protected regions.

Abstract

Supporting error resilience in future exascale-class supercomputing systems is a critical challenge. Due to transistor scaling trends and increasing memory density, scientific simulations are expected to experience more interruptions caused by transient errors in the system memory. Existing hardware-based detection and recovery techniques will be inadequate to manage the presence of high memory fault rates. In this paper we propose a partial memory protection scheme based on region-based memory management. We define the concept of regions called havens that provide fault protection for program objects. We provide reliability for the regions through a software-based parity protection mechanism. Our approach enables critical program objects to be placed in these havens. The fault coverage provided by our approach is application agnostic, unlike algorithm-based fault tolerance techniques.