I85PREP
⌘ K
DCAT study materials
DCAT study resources

The Overloaded Rack Row

DCAT · Put your knowledge to work in context.

All case studies

The Overloaded Rack Row

You are the facilities engineer for a 40-rack row in a colocation hall. The row was originally designed for 4 kW per rack with a raised floor plenum of 18 inches and perforated tiles at roughly 30% open area. Over two years, tenants have added compute, and the row now averages 7 kW per rack, with three racks at 11 kW. The hall's computer room air handlers (CRAHs) are running at 85% fan capacity and supply air at 18°C. Tenants have filed three hot-aisle complaints in the last month, all in the middle third of the row. Return air temperatures at the CRAH are 29°C. The building cannot add CRAH units for at least six months, and the colo contract caps per-rack power at the nameplate value on each circuit.

1. Before recommending any hardware, what is the most likely physical cause of the mid-row hot spots, and how would you verify it?

Read the worked answer

The most likely cause is bypass airflow and poor cold-aisle containment combined with insufficient tile open area for the new heat density. At 7 kW average and 11 kW peaks, the original 30% perforated tiles cannot deliver the required cold-air volume, and without containment, hot exhaust recirculates into the cold aisle - worst in the middle of the row, where static pressure in the plenum is lowest (the ends sit closest to the CRAH discharge).

Verification steps:

  1. Measure plenum static pressure at several points along the row - expect a measurable drop toward mid-row.
  2. Thermal traverse: log inlet temperatures at the top, middle, and bottom of each rack's front face. A vertical gradient (top much warmer) indicates recirculation over the top; a horizontal gradient indicates tile starvation.
  3. Check tile airflow with a flow hood on a sample of tiles; compare actual CFM to the tile's rated open-area flow at the measured plenum pressure.
  4. Confirm CRAH return/supply delta (29°C return vs 18°C supply is an 11°C delta - reasonable, so the CRAHs are not the primary failure; distribution is).

Only after these measurements should you propose remedies (higher-open-area tiles, containment, blanking panels).

2. Given the six-month CRAH constraint, propose a prioritized remediation plan and justify the order.

Read the worked answer

Priority 1 - free or near-free airflow hygiene (days):

  • Install blanking panels in all unused U-space. Open RU gaps let hot exhaust short-circuit into the cold aisle.
  • Seal cable cutouts in the floor and any open tile gaps.
  • Re-balance perforated tile placement: move tiles away from over-provisioned locations toward the three 11 kW racks and the mid-row.

Priority 2 - increase cold-air delivery (weeks):

  • Swap 30% tiles for higher open-area tiles (e.g., 50–60%) in the hot zones. This raises flow without new CRAH capacity.
  • Consider variable-speed fan tuning if the CRAHs support it - 85% fan speed may be masking a control setpoint issue; verify the CRAHs are not fighting each other.

Priority 3 - containment (1–3 months):

  • Cold-aisle containment (or hot-aisle, whichever fits the hall) is the single highest-leverage fix because it eliminates recirculation regardless of tile flow. It also lets you raise supply temperature later for efficiency.

Priority 4 - capacity (6 months):

  • Add CRAH units and/or in-row cooling for the three 11 kW racks.

Justification for order: Steps 1–2 are low-cost, reversible, and address the measured symptoms immediately. Containment is the structural fix but takes procurement and downtime planning. New CRAH capacity is last because it is the most expensive and slowest, and the earlier steps may reduce the required added capacity.

3. A tenant asks to raise one circuit to 13 kW to add a GPU node. What contractual and engineering checks must happen before you say yes?

Read the worked answer

Contractual checks:

  1. The colo contract caps per-rack power at the nameplate value on each circuit. Adding load above nameplate requires a contract amendment, not just an engineering approval.
  2. Verify the circuit breaker rating and conductor ampacity support 13 kW continuous. At 208V single-phase, 13 kW ≈ 62.5 A; at 208V three-phase, ≈ 36 A. If the breaker is 30 A or 60 A, 13 kW may exceed the 80% continuous-load rule (NEC 80% derating).
  3. Confirm PDU and upstream panel capacity - the row's electrical distribution may not have headroom.

Engineering checks:

  1. Cooling capacity: the rack is already at 11 kW; 13 kW raises local heat density further. Confirm cold-aisle delivery and return path can handle it (likely not until containment is in).
  2. Redundancy impact: does the added load erode N+1 on the PDU or CRAH serving that rack?
  3. Weight and cabling: GPU nodes add weight; verify floor loading and cable management.

Answer: Approve only after the contract amendment, an electrical load calculation showing ≤80% of breaker rating, and a thermal verification (ideally post-containment). Offer a staged approval: temporary load with monitoring, then permanent once containment is installed.