Modeling and Characterizing Service Interference in Dynamic Infrastructures

Performance interference can occur when various services are executed over the same physical infrastructure in a cloud system. This lead to performance degradation compared execution of isolation. work proposes Confirmatory Factor Analysis (CFA)-based model estimate across containers, caused by use CPU, memory and IO number co-hosted applications. The approach provides resource characterization through human comprehensible indices expressed as time series, so entire lifetime service be analyzed. Our experiments, based on combination real with different profiles Docker suggest that our accurately predict overall time, for combinations. used designer identify phases , during life-cycle service, likely greater degree interference, ensure only complementary hosted machine. Interference-awareness this kind will enable more intelligent management scheduling systems, may dynamically modify decisions.