D3.3 Validation Report

RECAP's work package (WP) 3 targets requirements analysis, use case adaptation, and validation of the platform against requirements and use case features. Outcomes of the WP are presented in three deliverables. Among these deliverables, the first one (D3.1) primarily introduced the use case, its user stories, its problems and requirements, and drafted how to validate the RECAP solutions to the use cases' problems. Next, (D3.2) mainly refined the requirements and detailed the validation methodology and its implementation through a set of validation scenarios and corresponding experiments/tests for each of the use cases. In the last stage of RECAP, all the defined tests are carried out and test results are collected and reported in this document – the final deliverable (D3.3) of the WP.

The key contribution of the document is the demonstration of the validation experiments, what and how test results are collected and analyzed, and what conclusions are drawn for each of the use cases. The document complements (D3.2) which details the entire validation process. More specifically, details of or changes made on the experiment setup and configurations are provided in D3.3. For each test defined in (D3.2), based on the proposed method, all the required experiments are implemented to accomplish the test, given that a test requires different experiments or runs in some cases. Comparisons and assessment of the results and conclusions drawn based on given acceptance criteria are finally reported. The success and/or acceptance of the RECAP methods is provided by use case owners as a final conclusion for each use case.

Summarizing, RECAP can provide significant improvements for all use cases. In Network Function Virtualization, we could show that RECAP can save at least 24% of resources (and therefore costs) due to auto-scaling. Validation of the Data Analytics Engine application has led to a changed application architecture and deployment strategy that could improve QoS while reducing operational costs by 80%. Validation of the vCDN use case shows that an impact of up to 24% on network cost savings can be achieved through right cache placement. Traffic forecasting can be used to reduce typical resource under-utilisation from >50% to ~12% facilitating significant power savings. Finally, RECAP shows that cost minimums could be found for CDN saving up to 25% on compute and storage investments. These savings equate to several tens of millions of Euros.