Table of contents

Fault Tolerant Extensions to Charm++ and AMPI

Outline

Slide 3

Background

Slide 5

Slide 6

Slide 7

Design Overview

Slide 9

Slide 10

Slide 11

Performance improve with Load Balancing

Slide 13

Motivation

Requirements

Design

Message Logging Protocol

Message Logging..

Checkpoint Protocol

Slide 20

Parallel Restart

Present Status

Overhead

Slide 21

Recovery Performance

Slide 26

Slide 27