Position Statement: Supporting Coordinated Adaptation in Networked Systems

While adaptation is widely recognized as valuable, adaptations in most existing systems are limited to changing execution parameters in a single software module or on a single host. Our position is that the true potential of adaptation can only be realized if support is provided for more general solutions, including adaptations that span multiple hosts and multiple system components, and algorithmic adaptations that involve changing the underlying algorithms used by the system at runtime. Such a general solution must, however, address the difficult issues related to these types of adaptations. Adaptation by multiple related components, for example, must be coordinated so that these adaptations work together to implement consistent adaptation policies. Likewise, large-scale algorithmic adaptations need to be coordinated using graceful adaptation strategies in which as much normal processing as possible continues during the changeover. Here, we summarize our approach to addressing these problems in Cactus, a system for constructing highly-configurable distributed services and protocols [2]. When multiple related system components can adapt to changes in the system state, the adaptations performed by these components must be coordinated to achieve a consistent adaptation policy. To achieve this, we have implemented an adaptation controller architecture that is responsible for making adaptation decisions for related adaptive components. Adaptation policies are specified on a componentby-component basis using sets of fuzzy logic rules, and then composed along with rules to coordinate the actions of different components to form a single controller. The challenge, of course, is designing a set of fuzzy rules that reflect the best adaptation strategies for a given situation. Even when coordinated adaptation decisions are made, large-scale algorithmic adaptations still present a difficult challenge. Without special provisions, for example, an adaptive system may be unable to process normal application traffic while it is changing between different algorithms. To alleviate this problem, we have designed and implemented a graceful adaptation protocol that coordinates changes across