HPC out of the box. This pre-configured HPC solution is pre-tested and ready to run, making it ideal for labs that need serious compute power without the complexity of deployment, especially in the realm of AI infrastructure and high performance computing.
ICERACK is a pre-configured HPC appliance designed for research labs and AI startups that need serious compute infrastructure but don't have a dedicated HPC administrator to deploy and manage it.
It arrives racked, cabled, and configured. Warewulf 4 provisioning, Slurm job scheduler, BeeGFS parallel storage all installed, validated, and documented before it leaves our facility.
You plug it in, run a quick site validation with us on a video call, and submit your first job the same day.
Compute nodes (configuration varies by tier) are essential components of high performance computing and AI infrastructure. The setup includes a management node, which serves as the Slurm controller and provisioning server, alongside a storage node equipped with the BeeGFS parallel filesystem. Additionally, an InfiniBand switch meeting the HDR100 standard, with an NDR upgrade available, is part of the configuration. A management ethernet switch, along with all necessary cabling, PDUs, and rack accessories, completes this pre-configured HPC solution.
Rocky Linux 9 is deployed on all nodes, creating a robust AI infrastructure for high performance computing. Warewulf 4 enables stateless provisioning, allowing nodes to reimage with a single command. Slurm manages partitions, memory limits, cgroup enforcement, and email notifications effectively within this pre-configured HPC environment. BeeGFS offers parallel storage with a validated stripe configuration, ensuring efficient data handling. Lmod environment modules facilitate easy access to necessary software, while monitoring is enhanced with plain-language alerts.
The HPL benchmark run and efficiency have been documented for our AI infrastructure, showcasing the capabilities of our high performance computing system. STREAM memory bandwidth has been validated per node, and IOR storage throughput has been verified. Additionally, the InfiniBand fabric has been validated using ibdiagnet. An acceptance report has been compiled with all benchmark results, and an operations runbook has been written specifically for non-HPC admins utilizing our pre-configured HPC setup.
We offer comprehensive site installation support, whether remote or on-site, tailored for your AI infrastructure needs. Our half-day training for researchers and admin covers the essentials of high performance computing and the use of pre-configured HPC systems. Additionally, we provide job script templates for common workloads to streamline your processes. For the first 30 days, enjoy our hypercare service where we're available whenever issues arise. As an option, you can choose an ongoing managed ops retainer after the initial 30 days for continued support.
Compute nodes (configuration varies by tier) are essential components of high performance computing and AI infrastructure. The setup includes a management node, which serves as the Slurm controller and provisioning server, alongside a storage node equipped with the BeeGFS parallel filesystem. Additionally, an InfiniBand switch meeting the HDR100 standard, with an NDR upgrade available, is part of the configuration. A management ethernet switch, along with all necessary cabling, PDUs, and rack accessories, completes this pre-configured HPC solution.
Rocky Linux 9 is deployed on all nodes, creating a robust AI infrastructure for high performance computing. Warewulf 4 enables stateless provisioning, allowing nodes to reimage with a single command. Slurm manages partitions, memory limits, cgroup enforcement, and email notifications effectively within this pre-configured HPC environment. BeeGFS offers parallel storage with a validated stripe configuration, ensuring efficient data handling. Lmod environment modules facilitate easy access to necessary software, while monitoring is enhanced with plain-language alerts.
The HPL benchmark run and efficiency have been documented for our AI infrastructure, showcasing the capabilities of our high performance computing system. STREAM memory bandwidth has been validated per node, and IOR storage throughput has been verified. Additionally, the InfiniBand fabric has been validated using ibdiagnet. An acceptance report has been compiled with all benchmark results, and an operations runbook has been written specifically for non-HPC admins utilizing our pre-configured HPC setup.
We offer comprehensive site installation support, whether remote or on-site, tailored for your AI infrastructure needs. Our half-day training for researchers and admin covers the essentials of high performance computing and the use of pre-configured HPC systems. Additionally, we provide job script templates for common workloads to streamline your processes. For the first 30 days, enjoy our hypercare service where we're available whenever issues arise. As an option, you can choose an ongoing managed ops retainer after the initial 30 days for continued support.
HPC tender won, need to deliver. ICERACK provides a validated, documented high performance computing stack that you can deliver under your brand, featuring pre-configured HPC solutions with init6 as the technical backbone for your AI infrastructure.
Share your workload, team size, and any grant or budget constraints you have. We'll design the right configuration for your needs, whether it's AI infrastructure or a pre-configured HPC for high performance computing.
Copyright © 2025 init6 - All Rights Reserved.
We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.