Grant: Discovery Grants Program - Individual
University of British Columbia
$25,000.00
Building error-resilient applications on many-core platforms
FY 2011-12
Summary
Emerging applications demand increasing amounts of computational power, which cannot be satisfied by today's micro-processors. To keep up with this demand, many-core processors with tens or hundreds of cores will soon become a reality. However, many-core processors will experience high rates of hardware errors, due to the fundamental limits of physics and engineering. Traditional approaches that mask hardware errors from software incur high power-overheads, which limit their applicability in many-core processors. Therefore, future applications must be capable of anticipating and tolerating hardware errors in order to leverage the computational capacities of many-core platforms.
The proposed research program will investigate an application-centric approach to detect, diagnose and recover from errors in many-core processors. The detection approach uses compiler-based techniques to insert runtime checks into applications. The approach is completely automatic and requires no intervention from the application developer except for annotations on what data is critical to the application, i.e., important for the application's correctness. The diagnosis approach uses formal verification techniques to trace back the cause of failures and isolate the failure-causing errors. Finally, the recovery approach reconfigures the application to mitigate errors optimally based on online performance and power models. The main advantage of the proposed research program is that error detection, diagnosis and recovery are all performed entirely in software, which results in substantial power savings compared to traditional approaches.
The results of the proposed research program will enable a diverse range of applications from high-performance codes to desktop applications to execute reliably on many-core processors. This in turn will facilitate the rapid adoption of many-core processors, and accelerate progress in science and engineering.
Program
Discovery Grants Program - Individual
Recipient
Location
British Columbia, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
386463-2010
Name nom
Pattabiraman, Karthik
Department département
Electrical and Computer Engineering
Organization id
2
Institution établissement
University of British Columbia
Province en
British Columbia
Province fr
Colombie-Britannique
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2011
Competition year année de concours
2010
Award amount
25000
Program id
RGPIN
Program nane en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
Discovery grants programs
Group fr
Programmes de subventions à la découverte
Committee code
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
801
Area of application group en
Information and communication services
Area of application group fr
Services d'information et de télécommunications
Area of application en
Computer software
Area of application fr
Logiciel
Research subject code
2720
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Computer systems software
Research subject fr
Logiciel de systèmes informatiques
Partie
2
Nb partie
5
Application title
Building error-resilient applications on many-core platforms
Keyword
NSERC
Application summary
Emerging applications demand increasing amounts of computational power, which cannot be satisfied by today's micro-processors. To keep up with this demand, many-core processors with tens or hundreds of cores will soon become a reality. However, many-core processors will experience high rates of hardware errors, due to the fundamental limits of physics and engineering. Traditional approaches that mask hardware errors from software incur high power-overheads, which limit their applicability in many-core processors. Therefore, future applications must be capable of anticipating and tolerating hardware errors in order to leverage the computational capacities of many-core platforms.
The proposed research program will investigate an application-centric approach to detect, diagnose and recover from errors in many-core processors. The detection approach uses compiler-based techniques to insert runtime checks into applications. The approach is completely automatic and requires no intervention from the application developer except for annotations on what data is critical to the application, i.e., important for the application's correctness. The diagnosis approach uses formal verification techniques to trace back the cause of failures and isolate the failure-causing errors. Finally, the recovery approach reconfigures the application to mitigate errors optimally based on online performance and power models. The main advantage of the proposed research program is that error detection, diagnosis and recovery are all performed entirely in software, which results in substantial power savings compared to traditional approaches.
The results of the proposed research program will enable a diverse range of applications from high-performance codes to desktop applications to execute reliably on many-core processors. This in turn will facilitate the rapid adoption of many-core processors, and accelerate progress in science and engineering.