Case Studies

Compute platform for previsional algorithm processing

Background

Our client, for the definition of short, medium and long term energy production and distribution strategies, needs to process complex analysis and algorithms, written in multiple languages (Matlab, Python, R ) and with diverse execution times (from 10 seconds to 60 minutes).

The traditional model based on static servers has high operating costs, due to the need to quickly manage processing load peaks combined with low efficiency due to long periods of inactivity.

Decisions and Actions

Our proposal was to realize a centralized computing platform based on:

  • microservices application architecture

  • container based infrastructure orchestrated by Kubernetes

  • calculation modules based on serverless platform.

The project included:

  • first scouting, evaluation and benchmark of the available serverless computing platforms

  • building of a Proof of Concept, showing the end-to-end feasibility of solution and the compatibility with the company’s algorithms and infrastructure

  • the full platform development and deploy.

Results

The platform has brought the following benefits to the customer:

  • reduction of operating costs, due to the use of serverless modules that maximize the usage of the infrastructure

  • improvement of processing response times, as the code-based model allows to give priority to short-term calculations that are no longer blocked waiting for long-term executions

  • improvement of algorithm execution times, thanks to the dynamic allocation of resources

  • future extensibility to different languages, simply by instantiating dedicated serverless modules without impacting the existing platform.

Get in touch

Are you ready to make sense and make things?