With a Virtual Cluster from Thread Burst you get an on-demand HPC cluster with job scheduler and an optional login node.
Virtual Cluster Nodes are configured with 2×25Gb Ethernet for east-west traffic and 200Gb Infiniband or RoCE networks for application(MPI) traffic. Each cluster is isolated using a combination of ACLs and VXLANs or Infiniband partition keys (PKEY). Hardware accelerated IPSec is enabled for all East-West traffic and is available as an add-on for application traffic.
SLURM, the Simple Linux Utility for Resource Management, is the default job scheduler and resource manager for your virtual cluster. You can send jobs to your virtual cluster from a remote SLURM instance, an optional login Node, or by using our API.
Storage options include high-speed parallel filesystem (CEPH), pure NVMEoF, or flash-based object storage. Bulk storage tiers are also available.
Keep an eye on your Virtual Cluster from afar, get detailed hardware stats, OS stats, and logs through the web interface. Easily configure your Virtual Cluster to send logs and telemetry to external logging services.
Virtual Clusters are billed on a per-minute basis. You're only charged for the Nodes you reserve and it's easy to stop your cluster at any time to pause billing, you'll still be charged for storage space while your cluster is off.
We recognize that EDUs have many budgeting restraints, that is why we offer special academic pricing. Contact us and discover how affordable HPC can be.
Have complex security requirements? We offer Isolated Virtual Clusters that give you all the power and flexibility of a Virtual Cluster on the public cloud but with an isolated network complete with dedicated switches and support systems. Contact us for a brief consultation and see how and Isolated Virtual Cluster meets your security and performance needs.