Multidimensional cloud latency monitoring and evaluation

Measuring or evaluating performance of a Cloud service is a non-trivial and highly ambiguous task. We focus on Cloud-service latency from the user’s point of view, and, to this end, utilize the multidimensional latency measurements obtained using an in-house designed active-probing platform, CLAudit, deployed across PlanetLab and Microsoft Azure datacenters. The multiple geographic Vantage Points, multiple protocol layers and multiple datacenter locations of CLAudit measurements allow us to pinpoint with great precision if, where and what kind of a particular latency-generating event has happened. We analyze and interpret measurements over two one-month time-intervals, one in 2013 and one in 2016. As these traces are large, an automated interpretation has been appended to the data-capture process. In summary, we demonstrate the utility of the multidimensional approach and document the differences in the measured Cloud-services latency over time. Our measurements data is publicly available and we encourage the research community to use it for verification and further studies.