Publications

Abstract not provided.

Increased HPC capability comes with increased complexity, part counts, and fault occurrences. In- creasing the resilience of systems and applications to faults is a critical requirement facing the viability of exascale systems, as the overhead of traditional checkpoint/restart is projected to outweigh its bene ts due to fault rates outpacing I/O bandwidths. As faults occur and propagate throughout hardware and software layers, pervasive noti cation and handling mechanisms are necessary. This report describes an initial investigation of fault types and programming interfaces to mitigate them. Proof-of-concept APIs are presented for the frequent and important cases of memory errors and node failures, and a strategy proposed for lesystem failures. These involve changes to the operating system, runtime, I/O library, and application layers. While a single API for fault handling among hardware and OS and application system-wide remains elusive, the e ort increased our understanding of both the mountainous challenges and the promising trailheads. 3

Abstract not provided.

In the ACS Data Analytics Project (also known as 'YumYum'), a supercomputer is modeled as a graph of components and dependencies, jobs and faults are simulated, and component fault rates are estimated using the graph structure and job pass/fail outcomes. This report documents the successful completion of all SNL deliverables and tasks, describes the software written by SNL for the project, and presents the data it generates. Readers should understand what the software tools are, how they fit together, and how to use them to reproduce the presented data and additional experiments as desired. The SNL YumYum tools provide the novel simulation and inference capabilities desired by ACS. SNL also developed and implemented a new algorithm, which provides faster estimates, at finer component granularity, on arbitrary directed acyclic graphs.

Next-generation exascale systems, those capable of performing a quintillion (10{sup 18}) operations per second, are expected to be delivered in the next 8-10 years. These systems, which will be 1,000 times faster than current systems, will be of unprecedented scale. As these systems continue to grow in size, faults will become increasingly common, even over the course of small calculations. Therefore, issues such as fault tolerance and reliability will limit application scalability. Current techniques to ensure progress across faults like checkpoint/restart, the dominant fault tolerance mechanism for the last 25 years, are increasingly problematic at the scales of future systems due to their excessive overheads. In this work, we evaluate a number of techniques to decrease the overhead of checkpoint/restart and keep this method viable for future exascale systems. More specifically, this work evaluates state-machine replication to dramatically increase the checkpoint interval (the time between successive checkpoint) and hash-based, probabilistic incremental checkpointing using graphics processing units to decrease the checkpoint commit time (the time to save one checkpoint). Using a combination of empirical analysis, modeling, and simulation, we study the costs and benefits of these approaches on a wide range of parameters. These results, which cover of number of high-performance computing capability workloads, different failure distributions, hardware mean time to failures, and I/O bandwidths, show the potential benefits of these techniques for meeting the reliability demands of future exascale platforms.

Despite its seemingly nonsensical cost, we show through modeling and simulation that redundant computation merits full consideration as a resilience strategy for next-generation systems. Without revolutionary breakthroughs in failure rates, part counts, or stable-storage bandwidths, it has been shown that the utility of Exascale systems will be crushed by the overheads of traditional checkpoint/restart mechanisms. Alternate resilience strategies must be considered, and redundancy is a proven unrivaled approach in many domains. We develop a distribution-independent model for job interrupts on systems of arbitrary redundancy, adapt Daly’s model for total application runtime, and find that his estimate for optimal checkpoint interval remains valid for redundant systems. We then identify conditions where redundancy is more cost effective than non-redundancy. These are done in the context of the number one supercomputers of the last decade, showing that thorough consideration of redundant computation is timely - if not overdue.

As High-End Computing machines continue to grow in size, issues such as fault tolerance and reliability limit application scalability. Current techniques to ensure progress across faults, like checkpoint-restart, are unsuitable at these scale due to excessive overheads predicted to more than double an applications time to solution. Redundant computation, long used in distributed and mission critical systems, has been suggested as an alternative to checkpoint-restart on its own. In this paper we describe the rMPI library which enables portable and transparent redundant computation for MPI applications. We detail the design of the library as well as two replica consistency protocols, outline the overheads of this library at scale on a number of real-world applications, and finally outline the significant increase in an applications time to solution at extreme scale as well as show the scenarios in which redundant computation makes sense.

Abstract not provided.

Exascale systems will have hundred thousands of compute nodes and millions of components which increases the likelihood of faults. Today, applications use checkpoint/restart to recover from these faults. Even under ideal conditions, applications running on more than 50,000 nodes will spend more than half of their total running time saving checkpoints, restarting, and redoing work that was lost. Redundant computing is a method that allows an application to continue working even when failures occur. Instead of each failure causing an application interrupt, multiple failures can be absorbed by the application until redundancy is exhausted. In this paper we present a method to analyze the benefits of redundant computing, present simulation results of the cost, and compare it to other proposed methods for fault resilience.

For future parallel-computing systems with as few as twenty-thousand nodes we propose redundant computing to reduce the number of application interrupts. The frequency of faults in exascale systems will be so high that traditional checkpoint/restart methods will break down. Applications will experience interruptions so often that they will spend more time restarting and recovering lost work, than computing the solution. We show that redundant computation at large scale can be cost effective and allows applications to complete their work in significantly less wall-clock time. On truly large systems, redundant computing can increase system throughput by an order of magnitude. © 2010 IEEE.

Abstract not provided.

Abstract not provided.

Petaflops systems will have tens to hundreds of thousands of compute nodes which increases the likelihood of faults. Applications use checkpoint/restart to recover from these faults, but even under ideal conditions, applications running on more than 30,000 nodes will likely spend more than half of their total run time saving checkpoints, restarting, and redoing work that was lost. We created a library that performs redundant computations on additional nodes allocated to the application. An active node and its redundant partner form a node bundle which will only fail, and cause an application restart, when both nodes in the bundle fail. The goal of this library is to learn whether this can be done entirely at the user level, what requirements this library places on a Reliability, Availability, and Serviceability (RAS) system, and what its impact on performance and run time is. We find that our redundant MPI layer library imposes a relatively modest performance penalty for applications, but that it greatly reduces the number of applications interrupts. This reduction in interrupts leads to huge savings in restart and rework time. For large-scale applications the savings compensate for the performance loss and the additional nodes required for redundant computations.

Abstract not provided.

Abstract not provided.