Title :
The availability of cloud-based services: Is it living up to its promise?
Author_Institution :
Dept. of Comput. Sci. & Civil Eng., Univ. di Roma “Tor Vergata”, Rome, Italy
Abstract :
Cloud-based services have spread among customers, boasting a number of advantages over centralized and self-owned infrastructures. Clouds providers claim a higher resilience, due to the use of a large distributed array of servers and storage facilities. However, assessments of their actual availability are missing. Here an early attempt to provide a third-party assessment is presented. Though not disposing of first-hand measurements, we have collected data on outages from a number of sources. We use those data to evaluate the frequency and duration of outages. The average duration of outages appears to lie in the 8-10 hours range, with a significant dispersion. A statistical analysis through QQ-plots and mean excess plots shows that the tail of the distribution of outage durations is well modelled by a Generalized Pareto distribution.
Keywords :
Pareto distribution; cloud computing; statistical analysis; QQ-plot; cloud-based service; generalized Pareto distribution; mean excess plot; statistical analysis; third-party assessment; Availability; Cloud computing; Companies; Google; Servers; Shape;
Conference_Titel :
Design of Reliable Communication Networks (DRCN), 2013 9th International Conference on the
Conference_Location :
Budapest
Print_ISBN :
978-1-4799-0049-7