Back-of-the-Envelope Estimation

// Estimation framework as code
interface EstimationProblem {
  // Step 1: Define the question clearly
  question: string;

  // Step 2: Identify the key variables
  assumptions: Record<string, number>;

  // Step 3: Calculate step by step
  calculate(): EstimationResult;
}

// Example: Estimate Twitter's storage needs for tweets
class TwitterStorageEstimation implements EstimationProblem {
  question = "How much storage does Twitter need for tweets per day?";

  assumptions = {
    dailyActiveUsers: 300_000_000,       // 300M DAU
    tweetsPerUserPerDay: 0.5,            // Not everyone tweets daily
    avgTweetSizeBytes: 250,              // Text content
    metadataPerTweetBytes: 200,          // User ID, timestamp, indexes
    mediaAttachmentRate: 0.2,            // 20% of tweets have media
    avgMediaSizeBytes: 200_000,          // 200KB per image
  };

  calculate() {
    const a = this.assumptions;

    const totalTweetsPerDay = a.dailyActiveUsers * a.tweetsPerUserPerDay;
    // = 300M * 0.5 = 150M tweets/day

    const textStoragePerDay = totalTweetsPerDay * (a.avgTweetSizeBytes + a.metadataPerTweetBytes);
    // = 150M * 450 bytes = 67.5 GB/day

    const mediaStoragePerDay = totalTweetsPerDay * a.mediaAttachmentRate * a.avgMediaSizeBytes;
    // = 150M * 0.2 * 200KB = 6 TB/day

    const totalPerDay = textStoragePerDay + mediaStoragePerDay;
    // = ~6 TB/day (media dominates)

    const totalPerYear = totalPerDay * 365;
    // = ~2.2 PB/year

    return {
      tweetsPerDay: "150 million",
      textStoragePerDay: "~67.5 GB",
      mediaStoragePerDay: "~6 TB",
      totalPerDay: "~6 TB",
      totalPerYear: "~2.2 PB",
      conclusion: "Media storage dominates. Need distributed object storage like S3.",
    };
  }
}