17-811 Self-Healing Systems: Class Discussion Summary

Summary of Class Discussion for April 7 by David Garlan

Selected notes that complement the summary written by Owen.

1. Rainbow versus Kramer versus Aura: One way of comparing these systems is in terms of the placement of abstract models and the parts of the system that translate between system and model

• Rainbow: externalized model; externalized translators (abstractors and refiners)
• Kramer: localized models; localized translators
• Aura: externalized model; localized translator

2. Joao’s layered depiction

• Issue 1: With respect to monitoring, how do we represent the environment in Aura self-healing architecture? I.e., what produces the M of the environment? A pseudo-component?
• Issue 2: Where does context observation fit into the model? Is that really part of the M of the environment?
• Issue 3: Fourth column could represent “scope”: higher levels are “aware” of more global perspectives.
• Issue 4: Is the M of one layer just the A of the next? If the contents of A is merely advisory (hints), then it may make sense to unify these concepts. Does it make sense in general? Important corrolary: In Aura control is peer-to-peer; abstraction induces layers