The capstone brief, team and design process

This capstone brings together everything in the data centre engineering program. Working as a design team, you will take a data centre campus from a client's first brief to a coordinated concept design with a cost estimate, schedule and final presentation, just as a consulting engineering firm or a design-build team would in practice. Real data centre projects are won or lost on integration: an elegant cooling design is worthless if the electrical system cannot power it, the structure cannot carry it, the fire protection does not suit it, or the budget cannot afford it. The capstone is therefore organized around a single project that each lesson develops further, so that decisions made early, such as rack density, redundancy level and site, flow through every later discipline. This first lesson introduces the project brief, the roles in a design team, the stages of design and their deliverables, the basis of design document, requirements management, and the capacity planning that sets the size and phasing of the campus.

The project: Aurora Campus

The client, a colocation and cloud services provider, wants to build a new campus called Aurora. The brief, summarized from a request for proposal, states:

• Ultimate capacity: 48 MW of critical IT load, built in two phases of 24 MW each, with land and utility capacity reserved for a possible third phase.
• Customers: a mix of enterprise colocation (average 10 to 15 kW per rack, air cooled) and AI training and inference clusters (40 to 130 kW per rack, liquid cooled), with the split uncertain and expected to shift toward AI over time.
• Availability: concurrently maintainable infrastructure for all critical systems, with the option of fault-tolerant power for selected halls.
• Efficiency and sustainability: annualized PUE of 1.25 or lower at full load, water use minimized, and a plan for 100 % carbon-free electricity matching on an annual basis, with a goal of hourly matching later.
• Schedule: phase 1 ready for customer fit-out within 30 months of site acquisition.
• Budget: a target capital cost to be confirmed by the team's estimate, benchmarked against market costs per megawatt.
• Location: a site in Canada to be selected by the team from three candidates, with access to a provincial transmission network and long-haul fibre.

Briefs are rarely complete or internally consistent. Part of the design team's job is to identify gaps and conflicts, such as a low PUE target combined with a requirement for minimal water use in a hot climate, and resolve them with the client early.

The design team

A data centre design team brings together many disciplines, coordinated by a design lead or project manager:

• Design lead or project manager: owns the schedule, budget, client relationship and integration across disciplines.
• Electrical engineering: utility interconnection, medium- and low-voltage distribution, UPS, generators, grounding, lighting, and power monitoring.
• Mechanical engineering: cooling plant, air and liquid cooling distribution, heat rejection, ventilation, plumbing and controls.
• Civil and structural engineering: site grading, stormwater, roads, foundations, structural frame, floor loading and seismic design.
• Architecture: building layout, envelope, security zoning, code compliance, and accessibility.
• Fire protection: detection, suppression, life safety and fire separations.
• Telecommunications and network: meet-me rooms, pathways, cabling plant and network architecture.
• Specialists: acoustic, environmental, security, commissioning, cost estimating, scheduling and sustainability consultants.

In the capstone, each team member takes primary responsibility for one or two disciplines and acts as reviewer for another. Weekly coordination meetings, a shared model or drawing set, and a clear log of decisions and open issues keep the work integrated.

Design stages and deliverables

Building design typically proceeds through stages, each ending with a deliverable that the client reviews and approves before the next begins:

The capstone targets the concept or schematic design stage: enough engineering to show that the design works, fits the site and budget, and meets the brief, with the main systems sized and arranged, but without the full detail of construction documents. Cost estimate classes follow the AACE International classification, in which a Class 4 estimate has a typical accuracy range of roughly −15 % to −30 % on the low side and +20 % to +50 % on the high side.

The basis of design

The basis of design (BOD) is the central document of the concept stage. It records the requirements and assumptions behind every major decision so that all disciplines, the client, and later the contractor and commissioning agent work from the same foundation. A data centre BOD typically includes:

• Project description, phasing and capacity.
• Design criteria: IT load per hall and per rack, redundancy level for each system, availability targets, environmental conditions (temperature and humidity ranges for air-cooled IT, supply temperatures for liquid cooling), climate design data, and efficiency targets.
• Codes, standards and guidelines that apply, such as the building and fire codes, electrical code, ASHRAE thermal guidelines, NFPA 75 and 76, and industry frameworks such as the Uptime Institute tiers or ANSI/TIA-942.
• System descriptions for each discipline, with sequences of operation at a conceptual level.
• Assumptions and open questions to be resolved with the client.

Closely related is the owner's project requirements (OPR) document, which states what the owner needs in their own terms. The BOD shows how the design will meet the OPR; commissioning later verifies both.

In Canada, the BOD should identify the editions of the National Building Code (as adopted by the province, for example the Alberta Building Code or Ontario Building Code), the Canadian Electrical Code Part I as adopted provincially, and the National Fire Code that apply on the selected site.

Requirements management

Large designs involve hundreds of requirements, from the client's availability target to a specific customer's rack density. Requirements management keeps them visible and traceable. Each requirement gets an identifier, a source, a responsible discipline and a verification method (design calculation, model, test or commissioning). A traceability matrix links requirements to design decisions and to the tests that will confirm them. When a requirement changes, as rack densities often do during data centre projects, the matrix shows which parts of the design are affected.

Changes are inevitable, so the team needs a simple change control process: a change request describes the change and its reason, each discipline assesses the impact on design, cost and schedule, the design lead and client decide, and the decision is recorded. Uncontrolled changes, accepted informally in meetings, are a major cause of coordination errors and cost overruns.

Capacity planning and phasing

The capacity plan determines how much infrastructure to build and when. Building everything at once wastes capital on idle equipment and risks building the wrong thing if customer needs change; building too little risks losing customers when capacity runs out. Phased designs build the shell, site infrastructure and utility connections for the ultimate capacity where that is economical, and add power and cooling equipment in blocks as demand grows. A key question is when to start each new phase, given that design, procurement of long-lead equipment such as transformers and generators, and construction can take well over a year.

# Capacity ramp for a two-phase campus: when must phase 2 construction start?
phase_mw = 24.0           # critical IT capacity per phase, MW
trigger = 0.80            # start phase 2 when forecast load reaches 80 % of phase 1
lead_quarters = 7         # design, procurement and construction lead time for phase 2
ramp = [1.5, 2.5, 3.0, 3.0, 2.5, 2.5, 2.0, 2.0, 1.5, 1.5, 1.5, 1.5, 1.0, 1.0, 1.0, 1.0]  # MW added per quarter

load = 0.0
trigger_q = None
print("quarter  IT load MW  utilization of phase 1")
for q, add in enumerate(ramp, start=1):
    load += add
    util = load / phase_mw
    flag = ""
    if trigger_q is None and util >= trigger:
        trigger_q = q
        flag = "  <- reaches 80 %"
    print(f"Q{q:<7} {load:9.1f}  {util:8.0%}{flag}")

full_q = next(q for q in range(1, len(ramp) + 1) if sum(ramp[:q]) >= phase_mw)
start_q = full_q - lead_quarters
print(f"phase 1 full in Q{full_q}; phase 2 must start by Q{start_q} to be ready in time")
print(f"the 80 % trigger in Q{trigger_q} leaves {full_q - trigger_q} quarters, short of the {lead_quarters}-quarter lead time" if full_q - trigger_q < lead_quarters else "trigger leaves enough lead time")

quarter  IT load MW  utilization of phase 1
Q1             1.5        6%
Q2             4.0       17%
Q3             7.0       29%
Q4            10.0       42%
Q5            12.5       52%
Q6            15.0       62%
Q7            17.0       71%
Q8            19.0       79%
Q9            20.5       85%  <- reaches 80 %
Q10           22.0       92%
Q11           23.5       98%
Q12           25.0      104%
Q13           26.0      108%
Q14           27.0      112%
Q15           28.0      117%
Q16           29.0      121%
phase 1 full in Q12; phase 2 must start by Q5 to be ready in time
the 80 % trigger in Q9 leaves 3 quarters, short of the 7-quarter lead time

The simulation shows a common trap. A simple rule of starting the next phase when the current one reaches 80 % utilization leaves only three quarters before phase 1 is full, while phase 2 needs seven. To avoid running out of capacity, the client must commit to phase 2 by quarter 5, when phase 1 is only about half full. In practice, owners manage this by pre-ordering long-lead equipment, designing phase 2 in parallel with phase 1 construction, and using customer pre-leasing commitments to justify early investment. The capstone design should show the phasing explicitly: what is built in phase 1, what is pre-built for later phases (such as underground ducts and utility substation bays), and what is added later.

Capstone deliverables and assessment

By the end of the course, each team will submit:

• A basis of design document covering all disciplines.
• A site selection report with scoring and justification (Lesson 2) and a permitting plan (Lesson 3).
• A capacity and load model (Lesson 4).
• Electrical concept: one-line diagrams, topology, equipment schedules and reliability analysis (Lesson 5).
• Mechanical concept: cooling strategy, plant sizing, water and energy analysis (Lesson 6).
• Civil, structural and architectural concept: site plan, building layout, structural criteria (Lesson 7).
• Fire protection and life safety concept (Lesson 8).
• Network and cabling concept (Lesson 9).
• Controls, security and operations concept (Lesson 10).
• Cost estimate, schedule and risk register (Lesson 11).
• Sustainability assessment, design review responses and a final presentation (Lesson 12).

Assessment rewards integration and justification more than volume. A shorter design in which every decision is explained, every discipline is coordinated, and trade-offs are presented honestly will score higher than a longer one with unexplained choices or conflicts between disciplines.