Data Center Design Execution Lead

Data Center Design Execution Lead

When to Use

• Define requirements for a new data hall, cage, or colo deployment
• Compare build vs colo vs cloud extension for a workload footprint
• Size power, cooling, and rack count for current and 3–5 year growth
• Review mechanical/electrical single-line diagrams and rack elevations
• Plan hot/cold aisle, containment, and cable pathways
• Specify network demarc, cross-connects, and latency to cloud regions
• Run design → procurement → install → commissioning → handoff
• Prepare acceptance tests and operations runbooks for facilities team

When NOT to Use

• Terraform, cloud networking, managed Kubernetes → infrastructure-engineer
• Helm, cluster add-ons, pod troubleshooting → cluster-deployment-engineer
• Application integration ADRs → senior-system-architecture
• Multi-team software program RAID → technical-program-manager
• SOC 2 control catalog without facility scope → compliance-engineer
• Physical security policy program only → cybersecurity, information-security-engineer
• Executive comms for launches → communication-lead
• Utilization, GPU supply forecast, consolidation, refresh → data-center-compute-supply-efficiency
• Multi-site roadmap, portfolio funding, steering → data-center-portfolio-planning-execution-lead
• Delivery schedule, rack-ready gates, construction RAID → senior-data-center-capacity-delivery-manager

Related skills

Core Workflows

1. Requirements and site strategy

Capture:

• Workload type (enterprise IT, HPC/GPU, edge)
• Availability target (tier intent, RTO/RPO for facility)
• Growth: racks, kW, network ports by year
• Constraints: geography, latency to users/cloud, sustainability, budget

Decide: own build, colo cage, or cloud-only with small edge.

See `references/facility_requirements.md`.

2. Power and cooling

• Design power chain: utility → ATS/STS → UPS → PDU → rack
• Size for peak kW/rack (air vs liquid, GPU trays)
• Cooling: CRAC/CRAH, containment, setpoints per ASHRAE class
• Target PUE and measurement points

See `references/power_cooling.md`.

3. Physical layout

• Rack rows, aisle width, floor load (lbs/sq ft)
• Hot/cold containment; overhead vs underfloor delivery
• Cable trays, fiber pathways, ladder rack
• Staging, spares, and restricted zones

See `references/physical_layout.md`.

4. Network and connectivity

• Meet-me room, carrier diversity, cross-connect model
• Core/distribution/toR switching for facility (distinct from app network design in cloud)
• Latency and bandwidth to primary cloud region for hybrid

See `references/network_connectivity.md`.

5. Execution and commissioning

See `references/execution_commissioning.md`.

6. Operations handoff

• DCIM asset records match labels and diagrams
• Environmental and power alarms to NOC
• Maintenance windows and vendor SLAs documented
• Runbooks: escort, smart hands, emergency power-down

See `references/operations_handoff.md`.

Output standards

• Capacity table: rack ID, kW design, weight, network Uplinks
• One-line diagram summary (power and cooling) with margins noted
• Commissioning checklist with pass/fail and owner
• Risks: single points of failure, lead-time items, permit dependencies

When to load references

• Site, tier, capacity → references/facility_requirements.md
• Power, UPS, cooling, PUE → references/power_cooling.md
• Racks, aisles, containment → references/physical_layout.md
• Carriers, MMR, hybrid links → references/network_connectivity.md
• Build phases and commissioning → references/execution_commissioning.md
• DCIM and ops → references/operations_handoff.md

Execution and commissioning

Table of contents

1. Phase plan 2. Long-lead tracking 3. Commissioning tests

Phase plan

Use technical-program-manager for multi-vendor RAID across civil, MEP, and IT trades.

Long-lead tracking

Typical long-lead (order early):

• Generators, switchgear, UPS modules
• Core switches, fiber panels
• Custom containment, CDUs for liquid cooling
• Permits and utility power upgrades

Weekly critical-path review until install complete.

Commissioning tests

Document results; attach to handoff package. Red-tag unresolved life-safety items before energizing.

Facility requirements

Table of contents

1. Decision matrix 2. Tier intent 3. Capacity model

Decision matrix

Use cloud for elastic burst; use facility for data residency, latency, or sustained high kW/rack.

Tier intent

Align business RTO/RPO with facility design (Uptime Institute tier concepts, TIA-942):

Document actual design vs marketing tier claims—gap analysis for insurance and audits.

Capacity model

Spreadsheet rows per year:

Include growth headroom (20–30% power, 15% cooling) before ordering switch gear and PDUs.

Validate floor loading and ceiling height for cabinet depth and overhead cable tray.

Network and connectivity

Table of contents

1. Demarc and carriers 2. Fabric roles 3. Hybrid to cloud

Demarc and carriers

• MMR: carrier handoff; diverse entry paths where tier requires
• Cross-connects: colo order lead time; label both ends
• Diverse paths: separate conduits/building entries; document shared risk (single bridge, etc.)

Contract: SLA for repair, LOA/CFA process, demarc extension standards.

Fabric roles

Out-of-band management network: separate switches/VLAN; no shared fate with production under failure.

Software-defined overlays and cloud networking → infrastructure-engineer.

Hybrid to cloud

Document for each site:

• Primary cloud region and dedicated link (DX, ExpressRoute, Partner Interconnect, etc.)
• Bandwidth, latency RTT, failover path
• Encryption (MACsec, IPsec) and ownership

Latency-sensitive apps: place in region or site within RTT budget.

Operations handoff

Table of contents

1. DCIM and monitoring 2. Runbooks 3. Maintenance

DCIM and monitoring

Minimum DCIM records per rack:

• Asset ID, U positions, kW nameplate, weight
• Power feed (A/B), PDU circuit
• Network ports (switch, port, VLAN intent link)

Monitor and alert:

• Inlet temperature, humidity per row
• PDU branch current vs threshold
• UPS battery health, fuel level (if gen)
• Leak detection (liquid cooling)

Integrate to NOC; severity aligned with incident-management-engineer if shared ops model.

Runbooks

Maintenance

• Quarterly: torque checks, filter replacement, alarm test
• Annual: UPS battery test, thermal imaging, failover drill
• Vendor SLAs: response time, parts depot, escalation

Facilities owns MEP; IT owns racks and cabling above PDU—RACI in handoff doc.

After handoff, K8s and app deploy → cluster-deployment-engineer.

Physical layout

Table of contents

1. Rack standards 2. Aisle and containment 3. Labeling

Rack standards

• 42U common; confirm depth (800–1200 mm) for GPU and cable bend radius
• Weight: populate bottom-heavy; check floor point loads vs slab rating
• PDU placement: vertical PDUs vs rack PDUs; leave service space
• Spares: 1–2 racks per hall for break/fix and staging

Aisle and containment

Minimum aisle width per fire code and cabinet door swing; keep escape paths clear.

Cable management: separate fiber and copper trays; maintain bend radius; color code A/B paths.

Labeling

Standardize before install:

• Rack: DC-HALL-ROW-RU (e.g. DC1-A04-R12)
• Ports: TIA-606 style; match DCIM and switch config
• Power: PDU circuit → rack feed documented in as-built

As-built drawings updated within 30 days of substantial completion.

Power and cooling

Table of contents

1. Power chain 2. Rack density 3. Cooling and PUE

Power chain

Utility → (gen optional) → ATS/STS → UPS → PDU (rack/row) → C13/C19 → IT load

Margin: Plan facility at 80% of nameplate for sustained IT load; burst for GPU per vendor guidance.

Rack density

Mismatch (high kW in low-density hall) causes thermal trips and stranded space.

Cooling and PUE

• PUE = total facility power / IT equipment power; measure at defined points consistently
• Setpoints: ASHRAE A1–A4 classes; wider bands save energy if hardware allows
• Containment reduces bypass air; blanking panels mandatory
• Humidity and filtration per equipment specs

Commissioning: heat load bank or staged IT rollout; trend inlet/outlet temps per rack row.