Open-Ended Design: Scope and Design a Data Center From a Vague Prompt

Read the full interview experience this question came from →

Quick Overview

A deliberately vague, open-ended prompt to design a data center, with no requirements given. Tests how you scope ambiguity into explicit requirements, then reason about site selection, power and backup, cooling, network fabric, redundancy, physical security, operations and capacity growth.

Open-Ended Design: Scope and Design a Data Center From a Vague Prompt

Company: Microsoft

Role: Software Engineer

Category: System Design

Difficulty: medium

Interview Round: Onsite

In this loop's behavioral round, the interviewer switched to a deliberately vague, open-ended design question: design a data center. No requirements, scale or constraints came with it, so deciding which problem you are solving is part of the exercise. Scope the problem into concrete requirements, then design it. ### Constraints and Clarifications - Nothing beyond the one-line prompt is given. Any capacity, availability or location figures you use are your own stated assumptions. - The prompt can be read as a physical facility or as the infrastructure inside one. Say which reading you take and why, and be ready to switch if the interviewer steers you. - Expect the interviewer to judge how you handle the ambiguity as well as the technical content. ### Clarifying Questions - Who is this data center for: an enterprise's own systems, a cloud provider's region, or a colocation business renting space to others? - What workloads will it run (general web services, storage, ML training) and how dense will the racks be? - How much capacity is needed at opening, and how much growth should the site allow for? - What availability is required, and does this site need to survive alone or does it sit alongside other sites? - Are there constraints on location, budget, timeline or sustainability targets? ### Part 1 — Scope the problem Turn the one-line prompt into a problem you can design. Ask the questions that matter most, state the assumptions you will proceed with, and set out the requirements those assumptions imply. ```hint Find the dominant constraint Ask which single quantity, once fixed, determines the size of almost everything else in a facility like this. ``` #### What This Part Should Cover - The clarifying questions that most change the design - Explicit assumptions about purpose, workload, capacity and availability - Requirements derived from those assumptions - A plan for how you will use the rest of the time ### Part 2 — Design the facility and its infrastructure Given your assumptions, design the data center: where it goes, how it gets and distributes power, how it removes heat, how servers are arranged and networked, and how it connects to the outside world. Use rough numbers to size the main systems. ```hint Follow the electricity Trace power from the grid to a server and back out as heat, and size each stage along the way. ``` #### What This Part Should Cover - Site selection criteria - Power chain and backup, with rough sizing - Cooling approach matched to rack density - Rack layout, network fabric and external connectivity ### Part 3 — Reliability, security and operations Explain how the data center keeps running through equipment failures and maintenance, how it is protected physically, how it is operated and monitored day to day, and how capacity grows over time. ```hint Pull the plug on one box Pick any single piece of equipment, imagine it failing during planned maintenance on its twin, and ask what keeps the servers running. ``` #### What This Part Should Cover - Redundancy levels and failure domains for power, cooling and network - Physical security and fire protection - Monitoring, maintenance and hardware lifecycle - Capacity planning and what a single site cannot protect against ### What a Strong Answer Covers - Structured scoping that turns a vague prompt into explicit, defensible assumptions - Power and cooling treated as the dominant design constraints, with rough sizing - A coherent network design inside the facility and out to the internet - Redundancy reasoning tied to failure domains and maintenance, with cost trade-offs - Operational realism: monitoring, security, growth and limits of a single site - Calm, organized communication under ambiguity ### Follow-up Questions - The interviewer says the site will mostly host dense ML training clusters. What changes? - How would you decide between building your own facility, leasing colocation space and using a public cloud? - How would you cut energy and water use without lowering availability? - How would you design three such sites so that losing an entire site causes no outage?

Overview: A deliberately vague, open-ended prompt to design a data center, with no requirements given. Tests how you scope ambiguity into explicit requirements, then reason about site selection, power and backup, cooling, network fabric, redundancy, physical security, operations and capacity growth.

Read the full Microsoft Software Engineer interview experience this question came from

Microsoft logo
Microsoft
Dec 20, 2025
mediumSoftware EngineerOnsiteSystem Design
0
0

In this loop's behavioral round, the interviewer switched to a deliberately vague, open-ended design question: design a data center. No requirements, scale or constraints came with it, so deciding which problem you are solving is part of the exercise. Scope the problem into concrete requirements, then design it.

Constraints and Clarifications

  • Nothing beyond the one-line prompt is given. Any capacity, availability or location figures you use are your own stated assumptions.
  • The prompt can be read as a physical facility or as the infrastructure inside one. Say which reading you take and why, and be ready to switch if the interviewer steers you.
  • Expect the interviewer to judge how you handle the ambiguity as well as the technical content.

Clarifying Questions Guidance

  • Who is this data center for: an enterprise's own systems, a cloud provider's region, or a colocation business renting space to others?
  • What workloads will it run (general web services, storage, ML training) and how dense will the racks be?
  • How much capacity is needed at opening, and how much growth should the site allow for?
  • What availability is required, and does this site need to survive alone or does it sit alongside other sites?
  • Are there constraints on location, budget, timeline or sustainability targets?

Part 1 — Scope the problem

Turn the one-line prompt into a problem you can design. Ask the questions that matter most, state the assumptions you will proceed with, and set out the requirements those assumptions imply.

What This Part Should Cover Guidance

  • The clarifying questions that most change the design
  • Explicit assumptions about purpose, workload, capacity and availability
  • Requirements derived from those assumptions
  • A plan for how you will use the rest of the time

Part 2 — Design the facility and its infrastructure

Given your assumptions, design the data center: where it goes, how it gets and distributes power, how it removes heat, how servers are arranged and networked, and how it connects to the outside world. Use rough numbers to size the main systems.

What This Part Should Cover Guidance

  • Site selection criteria
  • Power chain and backup, with rough sizing
  • Cooling approach matched to rack density
  • Rack layout, network fabric and external connectivity

Part 3 — Reliability, security and operations

Explain how the data center keeps running through equipment failures and maintenance, how it is protected physically, how it is operated and monitored day to day, and how capacity grows over time.

What This Part Should Cover Guidance

  • Redundancy levels and failure domains for power, cooling and network
  • Physical security and fire protection
  • Monitoring, maintenance and hardware lifecycle
  • Capacity planning and what a single site cannot protect against

What a Strong Answer Covers Guidance

  • Structured scoping that turns a vague prompt into explicit, defensible assumptions
  • Power and cooling treated as the dominant design constraints, with rough sizing
  • A coherent network design inside the facility and out to the internet
  • Redundancy reasoning tied to failure domains and maintenance, with cost trade-offs
  • Operational realism: monitoring, security, growth and limits of a single site
  • Calm, organized communication under ambiguity

Follow-up Questions Guidance

  • The interviewer says the site will mostly host dense ML training clusters. What changes?
  • How would you decide between building your own facility, leasing colocation space and using a public cloud?
  • How would you cut energy and water use without lowering availability?
  • How would you design three such sites so that losing an entire site causes no outage?