Configuration space in complex software products can be viewed as a linear combination of vectors where each coordinate represents a configuration parameter from either discrete or continuous range of values. This talk reveals how a vast amount of configuration space can be spanned using most influential vectors. The amount of combinations imply a simple metric used to determine the number of jobs which is limited by hardware resources and machine time used to run tests on chosen configurations. Most influential vectors are determined based on the configurations that are deployed and used by the clients.
For example, Linux kernel has more then 6000 CONFIG_* options, majority of them binary and some of them ternary, resulting thus in more then 2^6000 combination space. OpenStack configuration space is even bigger. But eventually is possible to test most important configurations using only several hundred jobs where each job is used for testing one given configuration.
Practical approach to enlarge tested configurations coverage without adding new hardware
Short Talk (20 Minutes)
Food & Swag Sponsors
Learn more about each of our Event Sponsors.