Checkpointing algorithms and fault prediction

TL;DR

This paper analyzes how fault prediction techniques influence checkpointing strategies, providing an optimal algorithm and checkpointing period to improve fault tolerance in large-scale systems.

Contribution

It introduces an extended analysis framework incorporating fault predictor recall and precision, and derives optimal checkpointing policies.

Findings

Optimal checkpointing period derived analytically.

Framework assesses impact of predictor recall and precision.

Provides guidelines for fault predictor performance at large scale.

Abstract

This paper deals with the impact of fault prediction techniques on checkpointing strategies. We extend the classical first-order analysis of Young and Daly in the presence of a fault prediction system, characterized by its recall and its precision. In this framework, we provide an optimal algorithm to decide when to take predictions into account, and we derive the optimal value of the checkpointing period. These results allow to analytically assess the key parameters that impact the performance of fault predictors at very large scale.