Compute

AI as a Service (AI cloud) in addition to all Serverius IT infrastructure services

The future of any modern IT infrastructure is a mixture of tailor-made datacenter services with locally hosted AI processing power. Datacenter services as a base, with additional DLU, CPU, GPU compute power which can automatically scale up or down. Therefore Serverius offers its clients a multi-million Euro AI hardware cluster under a “pay per use” model for everyone who doesn’t want to invest in their own AI hardware cluster.

fujitsu dedicated server

Clustered Deep & Machine Learning cloud service, based on the famous Tensorflow framework.

Currently, our R&D engineers are testing the first alpha version of an AI as a Service solution that makes use of the famous Tensorflow framework. It’s the first step to offering fully automated AI cloud processing power as a service. With Fujitsu’s upcoming new DLU accelerated hardware a new Deep Learning as a Service will be offered under a “pay per use” model. Machine learning, utilizing NVIDIA GPUs, will also be offered as a service.

Your data is key, your AI models are gold. Therefore you will need secure, nonshared, hardware storage to store and forward your AI data and its learning results.

For this new service, all data will be stored on a nonshared, dedicated cloud server. Every AI as a Service package will include a free dedicated server, where only you will have root access to. Storage deployment will be fully automated withing the AI as a Service process, therefore you do not need to have any server management skills themselves. But if you would like to use the server for other purposes, you can get full access by using your private SSH keys.

In most cases, one of these servers will offer more than enough storage capacity. But in some cases the training data and results will require more storage space and performance. If this is the case you can simply order more servers with this simple shopping card.

Serverius primarily uses Fujitsu servers because of their high quality and low energy consumption. And their new DLU processor offers, compared with others, amazing speed results. If needed, you can request hardware from other vendors with this hardware consultancy form.

Dedicated trust

All the AI infrastructure building blocks are based on dedicated hardware per client. This will guarantee the highest possible service performance and security which allows you to manage your own environment according to your personal needs. And of course, all services are provided within the safe borders of the Dutch Serverius datacenters.

You can use the AI compute service with or without other Serverius services. Depending on your needs and setup, a connection can be made using the Internet, VPN/GRE tunnels, and datacenter cross-connects. Heavy users can also use the SpeedIX, DCspine or any other inter-connect platform to connect to their AI processing farm.