VAIX – Fault Tolerant Infrastructure For 24/7 High Load Machine Learning Service
Introduction
Project Background VAIX.ai is an AI start-up that creates deeply integrated and powerful AI solutions for B2C businesses, with the vision of transforming e-commerce through personalization. VAIX offers highly customized recommender and classification products that help consumer brands to personalize every communication and offering to an individual customer. To achieve the unrivalled accuracy and relevance that makes their products so unique, VAIX uses Deep Learning models which require large volumes of data, GPU-level processing power performance with consistent availability and fault tolerance.
VAIX Request Our mission was to build a solution able to scale with the needs of the VAIX. It had to be cost-efficient without compromising security and fault-tolerance. 24/7 ‘up time’ was mandatory which meant it had to be maintained during any technical changes in the system, the data collecting process or a program crash.
The Challenge
The Solution
We managed to build a high-availability infrastructure using operating principles rigorously tested and implemented over two years by our R&D team. By applying these principles to VAIX’s need we were able to significantly reduce our client’s time-to-market and enhance product reliability while keeping costs low.
- Telegraf
- NodeExporter
- cAdvisor
- Grafana
- Zabbix
- Jenkins
- Gitlab ci
- NFS
- CEPH
- Datastore: elasticsearch
- UI: Kibana, Graylog
- AWS Route53
- NGINX
- Traefik
“The most exciting part of this particular project was the continuous headway we were able to make with the implementation of the technology pool. We didn’t just blindly follow trends. Instead, we looked for the optimal combination of different tech approaches. Our work on the VAIX project demonstrated once again that thoughtful, skilled and cooperative teams will always prove superior to any single technology solution”
Benefits
“To create deeply integrated and powerful AI solutions which transform our customers’ businesses, we require a reliable and fault tolerant infrastructure. With our partners at K&C, we have achieved this challenging goal for our high-load services. We rely on K&C to leverage the benefits of Docker, Rancher, AWS Route53 and S3, ELK, Ceph, Traefik, and other advanced tools to design cost-effective solutions for our tasks.”
Value Gained
Contact us
Tell us about your project and we’ll get in touch immediately.
Or you can contact us directly