Fig 5 - uploaded by Nane Kratzke
Content may be subject to copyright.
Launching and terminating times (Kubernetes): The Kubernetes cluster was composed of one master and five worker nodes. Data of single cloud experiments E1 and E2 (see Table 6) is presented.
Source publication
Cloud-native applications are often designed for only one specific cloud infrastructure or platform. The effort to port such kind of applications into a different cloud is usually a laborious one time exercise. Modern Cloud-native application architecture approaches make use of popular elastic container platforms (Apache Mesos, Kubernetes, Docker S...
Contexts in source publication
Context 1
... node type Worker node type AWS eu-west-1 m4.xlarge (4 vCPU) m4.large (2 vCPU) GCE europe-west1 n1-standard-4 (4 vCPU) n1-standard-2 (2 vCPU) Azure europewest Standard A3 (4 vCPU) Standard A2 (2 vCPU) OpenStack own datacenter m1.large (4 vCPU) m1.medium (2 vCPU) Table 6. Experiments: Single cloud experiments E1 and E2 were mainly used to measure infrastructure specific timings (see Figure 5) and have been repeated 10 times. Multi-cloud transfer experiments E3: Multi-cloud experiments E3 are used to demonstrate the transfer of elastic container platforms at runtime between di↵erent cloud service infrastructures and to get a better understanding about runtimes for these kind of operations (see Figure 8). ...
Context 2
... that is an operation which is done only one time while bootstrapping the cluster. The reader should compare these platform runtimes with infrastructure specific runtimes presented in Figure 5. Even on the fastest provider it took more than three minutes to launch a cluster. ...
Context 3
... interest- ing findings especially regarding software defined network aspects in multi-cloud scenarios and reactiveness of public cloud service infrastructures might be of interest for the reader. Figure 5 shows the results of the experiments E1 and E2. The reader might be surprised, that cloud infrastructure operations have a substantial variation in runtimes although similar resources were used for the presented experiments. ...
Context 4
... reader might be surprised, that cloud infrastructure operations have a substantial variation in runtimes although similar resources were used for the presented experiments. Figure 5 shows that a cluster launch on AWS takes only 3 minutes but can take up to 15 minutes on Azure (median values). The termination is even more worse. ...
Context 5
... shutdown nodes before attaching compensating nodes (in case of transferability scal- ing actions) is an obvious solution! But it is likely not ressource ecient -especially if we consider di↵erent timing behaviours of public cloud service providers (see Figure 5, 7, 8, and 9). To investigate resilient, resource and timing ecient execution strategies could be a fruitful research direction to optimize MAPE loops for transferability scenarios. ...
Similar publications
A data scientific process (e.g., Obtain, Scrub, Explore, Model, and iNterpret (OSEMN)) usually consists of different steps and can be understood as an umbrella for the combination of different most modern techniques and tools for the extraction of information and knowledge. When developing a suitable IT infrastructure for a self-service platform in...
