Workload Requirements
Document all workloads to be deployed
Application name, workload type (CPU-bound, memory-intensive, GPU-accelerated), and the performance requirements for each.
Determine CPU requirements per workload
Core count, clock speed requirements, and whether the workload benefits from more cores or higher per-core performance.
Determine memory requirements per workload
Working set size at peak load. Include headroom for OS, monitoring agents, and future growth.
Determine storage I/O requirements
IOPS, throughput, and latency requirements. Determine whether NVMe, SAS, or SATA is appropriate for each workload.
Verify ISV certifications
For applications with specific hardware certification requirements (SAP, Oracle, Microsoft), verify that the planned OEM and configuration are certified.
Document GPU requirements (if applicable)
Model size (determines VRAM requirement), training vs. inference, batch size, and throughput requirements.
Facility Readiness
Verify available rack space
Confirm that sufficient rack space is available for the planned deployment, including cable management and airflow requirements.
Verify available power capacity
Confirm that available power capacity (kW) is sufficient for the planned deployment at full load. Include redundancy requirements.
Verify cooling capacity
Confirm that cooling capacity is sufficient for the planned power density. GPU deployments above 10 kW/rack require liquid cooling verification.
Verify network port availability
Confirm that sufficient network ports are available at the required speed (25/100/200/400 GbE) for the planned deployment.
Verify out-of-band management network
Confirm that the out-of-band management network (iDRAC, iLO, XCC) is available and configured for the planned deployment.
Configuration Validation
Validate configuration against workload requirements
Confirm that the planned configuration (CPU, memory, storage, network) meets the documented workload requirements.
Validate power supply configuration
Confirm that power supplies are redundant and sized for the maximum configuration, not just the initial configuration.
Validate firmware and driver compatibility
Confirm that the planned firmware and driver versions are compatible with the operating system and applications to be deployed.
Validate BIOS/UEFI settings
Document the BIOS/UEFI settings required for the planned workloads (NUMA configuration, hyperthreading, power management).
Procurement Preparation
Confirm lead times
Verify lead times for all hardware components, particularly GPU hardware (12–26 weeks). Align procurement timeline with deployment requirements.
Verify OEM authorization
Confirm that the reseller is authorized by the OEM for the hardware being purchased. Unauthorized resellers do not provide OEM warranty coverage.
Confirm support contract terms
Verify response time, parts availability, and escalation path for the support contract. Confirm that the support level matches the criticality of the workloads.
Plan hardware disposition
Document the disposition plan for hardware being replaced. Certified data destruction is required for hardware that has processed regulated data.
Deployment Preparation
Plan staging and receiving
Identify the staging location where hardware will be received, inspected, and configured before deployment to the production environment.
Plan rack layout
Document the planned rack layout, including cable management, airflow requirements, and weight distribution.
Plan OS and application deployment
Document the OS installation and application deployment process. Automate where possible using PXE boot, Kickstart, or configuration management tools.
Plan integration testing
Define the integration tests that must pass before the new hardware is placed in production. Include performance benchmarks for the specific workloads.
Plan migration from legacy hardware
Document the migration sequence for workloads moving from legacy hardware. Include rollback procedures for each migration event.
GPU deployments require additional facility planning