We are in the middle of configuring a better access to our hpc-as-a-service to the public as there’s a huge demand of it. Like always, we want our clients to access the true bare-metal power of computing on the high-performance computer and not through virtualization. In one way, it’ll save our costs in term of manpower, in one way you’ll have the maximum power on it without sharing it with others.
We will also pre-installed AI frameworks like TensorFlow, PyTorch, queue managers etc upon request. So it will have the scalability in mind. By default also, we’re using InfiniBand 100Gbps FDR in all our nodes, and we shall upgrades those to 400Gbps in time.