Aviral Shrivastava: Teaching

Home

Publications

Teaching

Service

Lab

ASU 101

CSE 230

CSE 310

CSE 325

CSE 420

CSE 591 PEC

CSE 591 ARC

Multicore Programming

CSE 591: Advances in Reliable Computing

Course Objective: To develop a technique to improve processor reliability and submit the paper to an international conference

Course Abstract

As the technology scales and the transistor sizes diminish to provide us better performance and lower power, maintaining system reliability becomes a challenge. This course is about recent advances in reliable computing, and in particular we will focus on the increasing threat from soft errors, and the techniques researchers have developed to meet this challenge. This is a seminar plus project course, in which we will learn primarily through reading papers and through projects. In the class we will read papers, understand fault mechanisms, study typical fault models, and browse through the literature to see what mechanisms have researchers developed to make computation reliable. The second parallel part will be the project, in which you have to propose and implement your own technique for reliable computing. Your solutions can range from circuit level to system-level techniques to protect embedded to distributed supercomputing systems. You are also encouraged to come up with theoretical or analytical contributions.

Teaching Staff

Instructor: Aviral Shrivastava, aviral.shrivastava@asu.edu, 203-15 Centerpoint.

Class time

TT BYAC 270, 6:00 pm - 7:15 pm.

Office Hours

Aviral: TT 7:15 pm - 8:00 pm, MW 8:00 am - 9:00 am, 203-15 Centerpoint

Pre-requisites

Must have done computer architecture course equivalent to CSE 420.
Courses in operating systems, digital system design, and parallel programming are preferable.

Textbooks

Fault Tolerant Computer Architecture, Daniel Sorin, Morgan and Claypool
Architecture Design for Soft Errors, Shubu Mukherjee, Morgan Kaufman
Fault Tolerant Systems, Israel Koren and C. Mani Krishna

Course Contents

Motivation for reliable computing (1 week)
Techniques to protect caches (1 week)
Techniques to protect register files (1 week)
Redundant multicores (1 week)
Redundant multithreading (1 week)
In-application instruction duplication(1 week)
Control flow checking (1 weeks)
Symptom based error detection (1 week)
Algorithm based fault testing (1 week)
Protect exascale computing from soft errors(1 week)
Protect GPU computation from soft errors (1 week)

Course Structure

Classroom: This course is organized as a discussion course with each class meeting allocated to the discussion of a particular topic. Before each meeting, all students are expected to read 1-2 papers describing recent research on that topic.
Survey: You will have to do a survey on the topic that you pick up. You will have to read several papers on the topic of your interest, and present and write a survey on that.
Project: There will be one project in the course, with the objective to come up with, implement and test an existing technique for fault tolerance. It will be great, if you have some ideas on how to improve upon the technique, and you can demonstrate that.

Grading

Class participation and presentations: 30%
Survey: 40%
Project implementation and report: 30%

Reading List

T1: Core-level Redundancy

T2: Redundant multithreading

T3: In-application instruction duplication

T4: Control flow checking

T5: Symptom-based fault detection

T6: Algorithm-based fault tolerance

T7: Fault tolerance in graphics processing unit

T8: Fault tolerance at exascale

Teams and Topic Choices

T2: Jawad Rehman and Abhay Shankar Ravi: Redundant multithreading
T3: Kirash Rahbar and Kian Fakhri: In-application instruction duplication
T4: Matthew Mellott and Elizabeth Lee: Control flow checking
T5: Mohanlal Chauhan and Moslem Didehban: Symptom-based fault detection
T6: Harshal Bhavsar and Ke Shang: Algorithm based fault tolerance
T7: Patrick O' Connor and Amy Mazzola: Fault tolerance in GPUs
T8: John Dorry and Tyler Johnson: Fault tolerance at exascale

Class Schedule

Send Aviral, the summary of all the papers in your problem domain.Present the project, your progress, and the timeline in 5 slides. Send a first draft of the survey. This must be in latex, in the fomat described below.

Date	Class	Notes	Deadline
01/13/2015	Introduction by Aviral	slides
01/15/2015	Guest lecture by Dr. Jeyapaul	slides
01/20/2015	Class structure and deliverables
01/22/2015	Introduction to soft errors	slides	Send the name of your team members, and research topic choices to Aviral
01/27/2015	Core-level redundancy	paper
01/29/2015	Redundant multi-threading	paper	Send 4 papers that you will survey to Aviral
02/03/2015	In application instruction duplication	paper
02/05/2015	Control flow checking	paper
02/10/2015	Symptom-based fault detection	paper	Submit 1 page summary of 4 papers in your topic area.
02/12/2015	Algorithm-based fault tolerance	paper
02/17/2015	Fault tolerance in GPUs	paper
02/19/2015	GemV Tutorial		Send Aviral a mail indicating which paper/technique are you going to implement.
02/24/2015	Fault tolerance at exascale	paper	Send mail to Aviral indicating what software/hardware tools you will develop your technique on, e.g., gem5, llvm.
02/26/2015	Redundant multi-threading	paper
03/03/2015	In application instruction duplication	paper	Send Aviral, the problem definition of your survey, and the list of all the papers that will be reading and including in the survey.
03/05/2015	Control flow checking	paper
03/10/2015	Spring break
03/12/2015	Spring break
03/17/2015	Project Update and Multi-threading
03/19/2015	Project Update		Present the project, your progress, and the timeline in 5 slides. Send a first draft of the survey. This must be in latex, in the fomat described below.
03/24/2015	Project Update	HERMES paper
03/26/2015	Symptom-based fault detection	paper
03/31/2015	Algorithm-based fault tolerance	paper
04/02/2015	Fault tolerance in GPUs	paper
04/07/2015	Fault tolerance at exascale	paper
04/09/2015	Survey on Redundant Multithreading
04/14/2015	Survey on In application instruction duplication
04/16/2015	Survey on Control flow checking		Submit experimental results to Aviral
04/21/2015	Survey on Symptom-based fault detection
04/23/2015	Survey on Algorithm-based fault tolerance
04/28/2015	Survey on Fault tolerance in GPUs
04/30/2015	Survey on Fault tolerance at exascale

Survey/Paper Formatting

Paper must be written in ACM proceedings format, 9-point type, and may not exceed 8 pages (all inclusive). Word and LaTeX templates for this format are available here. Submissions must be in PDF, printable on US Letter sized paper.

Academic Integrity

Any incidence of cheating in this class will be severely dealt with. This applies to homework assignments and tests. The minimum penalty for cheating will be that the student will not obtain any credit for that particular assignment (This means that if in a test and/or assignment a student is found to have cheated, he/she will obtain zero in that test and/or assignment). Students are encouraged to discuss with others the materials covered in class. However students should not discuss problems in assignments/tests. One tends to get very suspicious if two identically wrong results show up in the homework assignment and/or tests.