Our CLOUD and AI solutions in ACCORDION project

Bluesoft R&D team accomplished new important tasks within Accordion Project:

1. Development of Test Execution Analyzer “TEA-app” – Machine Learning application for automatic tests based on neural networks
2. We found the solution on storage the excessive big files with Virtual Machines (VM’s) by placing them in a Azure Cloud
   

What’s the “TEA-APP”? And how it works ? 

Test Execution Analyzer is an automatic testing application based on machine learning algorithms. The major goal of this app is to analyze test logs from application testing results and learn to identify weather is the success or failure. Several appropriate models were researched and compared: TF-IDF, Word2Vec, Doc2Vec and Neural Networks. After evaluation of accuracy testing results Neural Networks methodology was chosen as the most accurate.

The model describe the step by step process of working application:

Application make an automated testing and measure predictions based on historical data.

Following graphs show the Neural Networks accuracy, precision and recall results after 20 cycles (epochs) of trained dataset:

In comparison the following graphs show results after 7 learning cycles (epochs) of dataset:

It means that  the most efficient way to get the best results are dataset training in 7 cycles.

The other important mechanism that was developed to ACCORDION is placement of Virtual Machines to Azure Storage and integrate it with the platform:

“ The problem we faced was that in addition to Docker images we need to process Virtual Machine (VM) images in our application. Docker images can be stored in Docker registry and this is our approach. There is no ready-made solution for Virtual Machines and that is why we had to create such a mechanism.

Because of the fact that such images can weigh a lot there are many problems. Starting with just the disk space, through the ability to quickly transfer them from the user to  ACCORDION EDGE Storage (AC.E.S.) -. Therefore we decided to use binary file storage offered by cloud providers. The choice fell on Azure Blob Storage as we are highly competent in implementation of AZURE CLOUD solutions.

ACCORDION Portal allow users to upload ready-made Virtual Machine images and place them in Blob Storage. They will be backed up to the whole application, so reusing the same image in different versions will not be a problem. At the time of creation a new version of the application, the user will have at his disposal all the images that were previously uploaded.”- said Mateusz Kamiński –  lead dev. in Bluesoft R&D team.

ACCORDION project is an  EU Research project focusing on Cloud/Edge infrastructure targeting Next Generation Applications such as Virtual and Augmented Reality in terms of  Low latency, High-bandwidth.

For more information please contact the author.