This paper evaluates the suitability of the MapReduce model for multi-core and multi-processor systems. MapReduce was created by Google for application development on data-centers with thousands of servers. It allows programmers to write functional-style code that is automatically parallelized and scheduled in a distributed system. We describe Phoenix, an implementation of MapReduce for shared-memory systems that includes a programming API and an efficient runtime system. The Phoenix runtime automatically manages thread creation, dynamic task scheduling, data partitioning, and fault tolerance across processor nodes. We study Phoenix with multi-core and symmetric multiprocessor systems and evaluate its performance potential and error recovery features. We also compare MapReduce code to code written in lower-level APIs such as P-threads. Overall, we establish that, given a careful implementation, MapReduce is a promising model for scalable performance on shared-memory systems with simple parallel code

/pdf/evaluating-mapreduce-for-multi-core-and-multiprocessor-1xfheav61u.pdf

Evaluating MapReduce for Multi-core and Multiprocessor Systems

Trends in terrestrial neutron-induced soft-error in SRAMs from a 250 nm to a 22 nm process are reviewed and predicted using the Monte-Carlo simulator CORIMS, which is validated to have less than 20% variations from experimental soft-error data on 180-130 nm SRAMs in a wide variety of neutron fields like field tests at low and high altitudes and accelerator tests in LANSCE, TSL, and CYRIC. The following results are obtained: 1) Soft-error rates per device in SRAMs will increase x6-7 from 130 nm to 22 nm process; 2) As SRAM is scaled down to a smaller size, soft-error rate is dominated more significantly by low-energy neutrons (<; 10 MeV); and 3) The area affected by one nuclear reaction spreads over 1 M bits and bit multiplicity of multi-cell upset become as high as 100 bits and more.

Impact of Scaling on Neutron-Induced Soft Error in SRAMs From a 250 nm to a 22 nm Design Rule

Deep learning neural networks (DNNs) have been successful in solving a wide range of machine learning problems. Specialized hardware accelerators have been proposed to accelerate the execution of DNN algorithms for high-performance and energy efficiency. Recently, they have been deployed in datacenters (potentially for business-critical or industrial applications) and safety-critical systems such as self-driving cars. Soft errors caused by high-energy particles have been increasing in hardware systems, and these can lead to catastrophic failures in DNN systems. Traditional methods for building resilient systems, e.g., Triple Modular Redundancy (TMR), are agnostic of the DNN algorithm and the DNN accelerator's architecture. Hence, these traditional resilience approaches incur high overheads, which makes them challenging to deploy. In this paper, we experimentally evaluate the resilience characteristics of DNN systems (i.e., DNN software running on specialized accelerators). We find that the error resilience of a DNN system depends on the data types, values, data reuses, and types of layers in the design. Based on our observations, we propose two efficient protection techniques for DNN systems.

Understanding error propagation in deep learning neural network (DNN) accelerators and applications

We present here a report produced by a workshop on 'Addressing failures in exascale computing' held in Park City, Utah, 4-11 August 2012. The charter of this workshop was to establish a common taxonomy about resilience across all the levels in a computing system, discuss existing knowledge on resilience across the various hardware and software layers of an exascale system, and build on those results, examining potential solutions from both a hardware and software perspective and focusing on a combined approach.

The workshop brought together participants with expertise in applications, system software, and hardware; they came from industry, government, and academia, and their interests ranged from theory to implementation. The combination allowed broad and comprehensive discussions and led to this document, which summarizes and builds on those discussions.

/pdf/addressing-failures-in-exascale-computing-1b4vxg0uhl.pdf

Addressing failures in exascale computing

This book provides a comprehensive description of the architetural techniques to tackle the soft error problem. It covers the new methodologies for quantitative analysis of soft errors as well as novel, cost-effective architectural techniques to mitigate them. To provide readers with a better grasp of the broader problem deffinition and solution space, this book also delves into the physics of soft errors and reviews current circuit and software mitigation techniques. 

TABLE OF CONTENTS
Chapter 1: Introduction 
Chapter 2: Device- and Circuit-Level Modeling, Measurement, and Mitigation 
Chapter 3: Architectural Vulnerability Analysis 
Chapter 4: Advanced Architectural Vulnerability Analysis 
Chapter 5: Error Coding Techniques 
Chapter 6: Fault Detection via Redundant Execution 
Chapter 7: Hardware Error Recovery 
Chapter 8: Software Detection and Recovery

* Provides the methodologies necessary to quantify the effect of radiation-induced soft errors as well as state-of-the-art techniques to protect against them

/pdf/architecture-design-for-soft-errors-3k5z0ocvev.pdf

Architecture Design for Soft Errors

Transient errors caused by terrestrial radiation pose a major barrier to robust system design. A system's susceptibility to such errors increases in advanced technologies, making the incorporation of effective protection mechanisms into chip designs essential. A new design paradigm reuses design-for-testability and debug resources to eliminate such errors.

https://www.eng.auburn.edu/~agrawvd/COURSE/E7770_Spr07/READ/01401773.pdf

Robust system design with built-in soft-error resilience

This paper presents a built-in soft error resilience (BISER) technique for correcting radiation-induced soft errors in latches and flip-flops. The presented error-correcting latch and flip-flop designs are power efficient, introduce minimal speed penalty, and employ reuse of on-chip scan design-for-testability and design-for-debug resources to minimize area overheads. Circuit simulations using a sub-90-nm technology show that the presented designs achieve more than a 20-fold reduction in cell-level soft error rate (SER). Fault injection experiments conducted on a microprocessor model further demonstrate that chip-level SER improvement is tunable by selective placement of the presented error-correcting designs. When coupled with error correction code to protect in-pipeline memories, the BISER flip-flop design improves chip-level SER by 10 times over an unprotected pipeline with the flip-flops contributing an extra 7-10.5% in power. When only soft errors in flips-flops are considered, the BISER technique improves chip-level SER by 10 times with an increased power of 10.3%. The error correction mechanism is configurable (i.e., can be turned on or off) which enables the use of the presented techniques for designs that can target multiple applications with a wide range of reliability requirements

Sequential Element Design With Built-In Soft Error Resilience

This work provides a comprehensive summary of radiation-induced soft error rate (SER) scaling trends of key CMOS bulk devices. Specifically we analyzed the SER per bit scaling trends of SRAMs, sequentials and static combinational logic. Our results show that for SRAMs the single-bit soft error rate continues to decrease whereas the multi-bit SER increases dramatically. While the total soft error rate of logic devices (sequentials and static combinational devices) has not changed significantly, a substantial increase in the susceptibility to alpha particles is observed. Finally, a novel methodology to extract one-dimensional cross sections of the collected charge distributions from measured multi-bit statistics is introduced.

Radiation-Induced Soft Error Rates of Advanced CMOS Bulk Devices

We report on measured radiation-induced soft error rates (SER) of memory and logic devices built in a 22 nm high-k metal gate bulk Tri-Gate technology. Our results demonstrate excellent single event upset (SEU) scaling benefits of tri-gate devices. For cosmic radiation, SEU SER reduction levels of the order of are observed relative to 32 nm planar devices, while for alpha-particles, the measured SEU SER benefit is in excess of . Similar improvements are observed for Tri-Gate combinational logic and memory array multi-cell upset (MCU) rates. Reduced SER (RSER) device SER performances (relative to standard, non -RSER devices) are on par or better than that of tested 32 nm planar devices. Finally, a novel, efficient SER reduction design called RTS is introduced.

Soft Error Susceptibilities of 22 nm Tri-Gate Devices

We present two techniques for correcting radiation-induced soft errors in combinational logic ? Error Correction using Duplication, and Error Correction using Time-Shifted Outputs. Simulation results show that both techniques reduce combinational logic soft error rate by more than an order of magnitude. Soft errors affecting sequential elements (latches and flip-flops) at combinational logic outputs are automatically corrected using these techniques.

N. Seifert

Papers

Robust system design with built-in soft-error resilience

Sequential Element Design With Built-In Soft Error Resilience

Radiation-Induced Soft Error Rates of Advanced CMOS Bulk Devices

Soft Error Susceptibilities of 22 nm Tri-Gate Devices

Combinational Logic Soft Error Correction