DeepSquare Grid
The network of HPC clusters forming the pool of compute required to process users’ workloads. The DeepSquare Grid is composed of clusters from various providers.
Speed up your projects with ease on DeepSquare – the platform crafted by builders, for builders.
The network of HPC clusters forming the pool of compute required to process users’ workloads. The DeepSquare Grid is composed of clusters from various providers.
The open-source Kubernetes-based infrastructure orchestrator that automates the process of deploying, managing and monitoring compute-optimized clusters from bare metal servers to VMs and containers.
The builders, developers, AI engineers, artists, blockchain enthusiasts, disruptors, innovators and visionaries. Join us in our mission to democratize access to HPC!
DeepSquare uses YAML for defining a job's workflow.
DeepSquare is able to compile and validate the YAML workflow into a sophisticated script
compatible with job schedulers. This approach facilitates the creation of intricate
Directed Acyclic Graphs (DAGs) and workflows in a straightforward and comprehensible
manner, ensuring not only simplicity and clarity but also promoting repeatability and
safety within a consistent framework.
Moreover, with the YAML language server, users can benefit from the auto-completion for a
better IDE experience.
DeepSquare deploys its Meta-Scheduler on the Blockchain for transparency, and employs scheduling algorithms that consider in various factors to optimize compute allocation for equitable distribution and fairness.
DeepSquare is able to run any OCI containers with seamless support parallization technologies (MPI, NCCL, ...), and hardware accelerators such as GPUs, Infiniband, and more.
DeepSquare jobs have external network access, enabling users to execute interactive tasks without compromising infrastructure provider security, thanks to the implementation of a zero-trust infrastructure and strict firewall requirements.
DeepSquare charges users based on the actual usage of compute resources down to the minute, and not periodically. This allows users to save on costs and only pay for what they use.
DeepSquare's sustainability strategy revolves around providing a readily deployable, eco-friendly configuration that minimizes the environmental footprint of compute resources. By integrating cutting-edge technologies and prioritizing renewable energy sources, DeepSquare aims to reduce electrical consumption while making high-performance processing accessible to small businesses, researchers, and academic institutions.
We advocate Immersion Cooling, a practice of placing electronic equipment into a thermally conductive, and electronically resistive liquid that is 1,200 times more efficient than air at capturing heat. The heat captured through immersion cooling can be exchanged in a simple, but efficient liquid-to-liquid heat exchanger. The heat then can directly be reused for heating purposes or industrial processes.