Estimate the Time for a 100 PB Coast-to-Coast Data Migration

Quick Overview

A Google product manager estimation question: how long would it take to move 100 petabytes of data from the US East Coast to the West Coast? It tests unit conversion, clearly labeled bandwidth assumptions, comparing a network transfer with a physical shipment, and a sanity-checked final estimate.

Estimate the Time for a 100 PB Coast-to-Coast Data Migration

Company: Google

Role: Product Manager

Category: Product / Decision Making

Difficulty: medium

Interview Round: Other

Estimate how long it would take to get 100 petabytes of data from the East Coast of the United States to the West Coast. ```hint Find the bottleneck Convert the volume into units you can divide by a throughput, then ask what actually limits throughput from end to end. Do not assume the first transfer channel that comes to mind is the only one available. ``` ### Clarifying Questions - Must the data travel over a network, or can physical media be shipped? - How much bandwidth is available between the two sites, and can this move use all of it without hurting production traffic? - Does the clock stop when the last byte arrives, or when the copy has been verified and is usable at the destination? - Is the data static, or is it still being written to while the move happens? ### What a Strong Answer Covers - Clarifying what counts as "moved" and which transfer channels are allowed - Correct unit conversion and a clearly labeled bandwidth assumption, with the time worked out for more than one link speed - Considering a physical shipment and identifying where its real bottleneck sits - Practical overheads: realistic link utilization, verification, failures, and data that changes during the move - A clear recommendation with a range, a sanity check, and the assumptions that would flip it ### Follow-up Questions - At what sustained link speed does a network transfer beat shipping drives? - How would you keep the West Coast copy consistent if the data keeps changing during the move? - What would you recommend if this had to happen every month instead of once?

Overview: A Google product manager estimation question: how long would it take to move 100 petabytes of data from the US East Coast to the West Coast? It tests unit conversion, clearly labeled bandwidth assumptions, comparing a network transfer with a physical shipment, and a sanity-checked final estimate.

Google logo
Google
Aug 7, 2024
mediumProduct ManagerOtherProduct / Decision Making
0
0

Estimate how long it would take to get 100 petabytes of data from the East Coast of the United States to the West Coast.

Clarifying Questions Guidance

  • Must the data travel over a network, or can physical media be shipped?
  • How much bandwidth is available between the two sites, and can this move use all of it without hurting production traffic?
  • Does the clock stop when the last byte arrives, or when the copy has been verified and is usable at the destination?
  • Is the data static, or is it still being written to while the move happens?

What a Strong Answer Covers Guidance

  • Clarifying what counts as "moved" and which transfer channels are allowed
  • Correct unit conversion and a clearly labeled bandwidth assumption, with the time worked out for more than one link speed
  • Considering a physical shipment and identifying where its real bottleneck sits
  • Practical overheads: realistic link utilization, verification, failures, and data that changes during the move
  • A clear recommendation with a range, a sanity check, and the assumptions that would flip it

Follow-up Questions Guidance

  • At what sustained link speed does a network transfer beat shipping drives?
  • How would you keep the West Coast copy consistent if the data keeps changing during the move?
  • What would you recommend if this had to happen every month instead of once?