Experience transfer for the configuration tuning of large scale computing systems
View Patent ↗A computer implemented method employing experience transfer to improve the efficiencies of an exemplary configuration tuning in computing systems. The method employs a Bayesian network guided tuning algorithm to discover the optimal configuration setting. After the tuning has been completed, a Bayesian network is obtained that records the parameter dependencies in the original system. Such parameter dependency knowledge has been successfully embedded to accelerate the configuration searches in other systems. Experimental results have demonstrated that with the help of transferred experiences we can achieve significant time savings for the configuration tuning task.
1. A computer implemented method for configuring and tuning computing systems comprising the computer implemented steps of:
determining a set of transferable experiences for a first computing system S 0 and representing that set of transferable experiences in a Bayesian network model;
extracting the set of transferable experiences from the first computing system S 0 during the modeling; and
embedding the extracted set of transferable experiences into a new computing system S 1 ;
wherein the transferable experiences reflect common characteristics of the two systems such that they remain valid after transfer to the new computing system, are related to a knowledge of managing the systems learned through the operation of the first computing system, and are expressed as hidden dependencies among system configuration parameters that cannot be manually specified by system operators.
2. The computer implemented method of claim 1 further comprising the step of:
tuning the configuration of the first computing system S 0 by performing a Bayesian network tuning configuration of the extracted set of transferable experiences prior to embedding wherein the Bayesian network is first constructed given a population of configuration samples and an output includes a network structure that represents the dependency relationships between configuration parameters as well as a joint distribution model that has been encoded by the dependency knowledge.
3. The computer implemented method of claim 2 wherein said embedded set of transferable experiences are represented as a Bayesian network.
4. The computer implemented method of claim 2 wherein the Bayesian network is incrementally constructed given a sequence population of configuration samples based on a minimum number of samples.