Self-Adapting Resource Bounded Distributed Computations
Nadeem Jamali, Xinghui Zhao · 2007
Self-adaptation is about computations adapting to their environments. The need for adaptation may dynamically arise as a result of evolving computations or the environment. An important part of the environment is the computational resources for which computations compete. The CyberOrgs model encapsulates distributed concurrent computations along with the computational and communication resources they require plus purchasing power for acquiring additional resources. Ownership of resources coupled with an effective control mechanism creates a predictable resource environment for computations to execute in - in a coordinated manner. CyberOrgs create three opportunities for self- adaptation: algorithms may be chosen using resource knowledge, additional resources may be purchased to adapt to evolving needs, and computations may coordinate use of known computational and network resources for optimal results. The CyberOrgs model is presented and a prototype implementation is described. Our experience with using CyberOrgs' resource awareness for hierarchical coordination of distributed processor resource delivery is presented. Experimental results show that resource knowledge based reasoning leads to efficient distributed adaptation.