Containment Domains : a Full System Approach to Computational

Future large-scale computer systems will require a comprehensive and flexible resilience solution to maintain correctness while satisfying strict energy constraints. Toward this end, this document presents the definition of the semantics of scalable and efficient containment domains (CDs) resilience framework. Containment domains are a programming construct that enables applications to express, tune, and specialize error detection, state preservation and restoration, and recovery to satisfy applicationspecific resilience needs. Containment domains have weak transactional semantics and are nested to take advantage of deep machine and application hierarchies. The flexibility and hierarchy of containment domains enables the programmer and software system to match state preservation, error detection, and recovery costs with the rate and severity of errors. Containment domains also empower the programmer to reason about resiliency and to utilize domain knowledge to improve efficiency beyond what compiler analysis can achieve without such information. This report describes CD semantics and identifies the system features and mechanisms required for an efficient containment domains implementation.