Distributed Email Service

Phase 1 — Clarify & Scope

Interviewer: Design a distributed email service — think Gmail or a large transactional email platform like SendGrid. What do you want to clarify?

Candidate: Key questions:

1. Are we building a user-facing email client (Gmail-style — read/compose/search), a transactional email sender (SendGrid-style — API to send emails at scale), or both?
2. What's the storage model — do we need to store received email, or just send?
3. What's the scale — users, messages per day?
4. What protocols — SMTP for sending, IMAP/POP3 for retrieval?
5. Search requirements — full-text search over email body?

Interviewer: Gmail-style — user email client with sending, receiving, storing, and searching. 1 billion users, 10 billion emails per day sent and received. Full-text search over emails. Standard protocols: SMTP in/out, IMAP for client access.

Candidate: Let me size this:

$\text{Emails/sec} = \frac{10 \times 10^9}{86{,}400} \approx 115{,}740 \text{ emails/sec}$

Average email size ~75 KB (headers + body + small attachment):

$\text{Storage/day} = 10^{10} \times 75 \text{ KB} = 750 \text{ TB/day}$

But most emails are stored for months/years. With 5-year retention:

$\text{Total storage} = 750 \text{ TB} \times 365 \times 5 \approx 1.37 \text{ EB (exabytes)}$

Each user has a 15 GB quota (like Gmail). Average utilization 20%:

$\text{Storage per user} = 15 \text{ GB} \times 0.2 = 3 \text{ GB average}$ $\text{Total} = 10^9 \times 3 \text{ GB} = 3 \text{ EB}$

Interviewer: What are the core requirements?

Candidate:

• Send email: compose → queue → deliver via…