Implementation of a cloud service for big data and AI
Project duration: 1 year, 5 months
Brief description
A new service is designed and built to extend the services of a data center. This service offers big data and AI-as-a-Service. The underlying platform is thereby based on dedicated hardware as well a suitable technology stack on top of IBM's Cloud Private.
Supplement
After evaluating multiple options for providing big data and AI as a service the choice has been made to build this on top of IBM Power PC and x86 hardware. The respective servers are build as a composite of management and worker nodes. These are supported by high performance GPU cards by Tesla to address AI specific calculation loads. Software-wise the respective services are supported by IBM's Cloud Private. This includes hosting of Docker containers, their management by Kubernetes and deployments by means of HELM charts.
Subject description
The development of methods and algorithms in support of artificial intelligence often requires extraordinary storage and computing resources. On one hand, large data sets are used for training and testing AI models. On the other hand, heavy computing loads are created, i.e. by neural networks. These calculation demands are often not met with suitable real-life performance by customary CPU's. Using high-power GPU's in support fills this gap to meet modern customer requirements.