Full stack system design

1.You are asked to design a ticket sale site for a stadium concert. Before drawing any boxes, which question shapes the design most?

2.An interviewer says, "Design a meal planning app," and you have 45 minutes. What is the strongest way to open?

3.Product says a clinic booking system must never double-book a slot. Which clarifying question matters most for that rule?

4.Stakeholders say a payroll export "must be real time." What should you ask before designing anything?

5.You are designing a code review tool. Which requirement changes the architecture the most?

6.A customer asks for 99.99 percent availability on a factory sensor dashboard. What should you clarify first?

7.A news feed has 5 million daily active users who each load about 20 feed pages a day. Peak traffic is 3 times the average. What peak request rate should you plan for?

8.A photo backup service has 200,000 users who each upload 5 photos a day averaging 2 MB. About how much new storage is needed per year before replication?

9.A live cooking class platform expects 10,000 concurrent viewers, each streaming at 3 megabits per second. What outbound bandwidth does that need?

10.A chat app expects 500,000 concurrent websocket connections, and each uses about 20 KB of server memory for buffers and state. How much memory do the connections need?

11.A shipping operations dashboard is open in 200,000 browser tabs, and each tab polls the API every 5 seconds. What does that imply?

12.A greenhouse monitoring system has 50,000 sensors, each sending a 100-byte reading every 10 seconds. What is the write rate and raw data per day?

13.A hardware store catalog has 10 million products. The hot 20 percent are cached as 2 KB JSON blobs. Ignoring overhead, how much cache memory is that?

14.A checkout API handles 2,000 requests per second, and each request holds a database connection for about 50 ms. How many connections are busy at once on average?

15.An API serves 3,000 requests per second on average, and each request writes about 2 KB of logs. About how much log data is that in a 30-day month?

16.A team messaging app has 1 million users who each send 40 messages a day averaging 200 bytes. How much message storage is that per year before indexes and replicas?

17.A recipe site has 2 million public recipe pages. Edits are rare and search ranking matters. Which rendering strategy fits best?

18.When a city marathon opens registration, 300,000 people keep refreshing the race page, which shows a spots-left banner. How should the page be served?

19.A local news site caches articles at its content delivery network (CDN), but the header shows the signed-in reader's name. How do you keep articles cacheable?

20.An online course catalog sets s-maxage=3600 at its content delivery network. Editors complain that fixes take an hour to appear. What is the best change?

21.A greenhouse analytics product has a dashboard behind login with highly interactive charts. Which rendering approach fits best?

22.A used car marketplace caches listing pages at its content delivery network, but the hit rate is 12 percent. URLs carry utm_source and other tracking parameters. What helps most?

23.A global content delivery network has 200 edge locations. After a site-wide purge, the origin gets flooded by misses from every edge. What helps most?

24.A community events site's origin goes down for 10 minutes. Most pages are cached at the content delivery network. How do you keep pages up during the outage?

25.The home screen of a grocery delivery app makes 6 REST calls in sequence, and it is slow on mobile networks. What do you propose?

26.A team proposes exposing its internal GraphQL schema as-is to outside partners of an invoicing product. What is the main concern?

27.You add an Idempotency-Key header to POST /payouts for a gig work platform. Two requests with the same key arrive 20 ms apart on different servers. What must the server do?

28.A spreadsheet-style inventory tool lets a user save 500 row edits at once. Which API design fits best?

29.A recipe platform's public search API gets bursts from one partner's batch jobs that hurt other partners. How should rate limiting work?

30.Mobile clients of a food truck ordering app stay on old versions for months. You need to change price from an integer to an object with a currency. What is the safest plan?

31.A wholesale order list endpoint returns full orders with line items, about 40 KB each, but the list screen shows only 4 fields. Mobile users say it is slow. What is the best fix?

32.A pharmacy delivery app shows the courier's location, updated every 5 seconds, flowing only from server to customer. Which transport fits best?

33.A shared whiteboard app sends cursor positions and strokes about 20 times a second in both directions. Which transport fits best?

34.A support chat app runs 20 websocket gateway nodes, and members of one conversation are connected to different nodes. How does a message reach all of them?

35.A deploy of a sports chat app's gateways drops 300,000 websocket connections. Every client reconnects at once and the auth service falls over. What is the fix?

36.A live cricket score feed uses server-sent events. A fan's phone loses signal in a tunnel for 30 seconds. How do you avoid gaps in the feed?

37.In a payroll tool, export jobs take 2 to 10 minutes, and about 50 users watch job status at any time. What is the simplest suitable design?

38.Server-sent event streams for a stock alert app get cut every 60 seconds when nothing is sent. What is the likely cause and fix?

39.Users of a family group chat see messages in different orders on different devices. What design gives one consistent order?

40.A social app shows online status for 2 million users, and each heartbeat writes a row to the main database. What is a better design?

41.A property management product has a web app, mobile apps, and a public partner API. Which authentication setup fits best?

42.A banking-style app uses 5-minute access tokens with refresh tokens. A user taps "log out everywhere." What makes that work?

43.How can a fitness tracking app detect that a refresh token has been stolen?

44.A wedding planning site keeps sessions in each server's memory with sticky load balancing. Every deploy logs everyone out. What is the fix?

45.A document editing product is split into several services behind an API gateway. Where should the check "can this user edit this document" live?

46.Support agents for a dental clinic booking system need to see exactly what a patient sees to debug issues. What is the right design?

47.A home video backup app uploads 4 GB files from phones on flaky networks. How should uploads work?

48.You issue presigned upload URLs for object storage in a scanned receipt app. What should each URL constrain?

49.Clients of a podcast hosting site upload audio straight to object storage, then call the API to attach it. Some never call back, and the API trusts whatever they send. What fixes both problems?

50.A greeting card site resizes user-uploaded images with a library that has had parsing security bugs. How should processing run?

51.A veterinary clinic stores medical records as files in object storage. How should pet owners download them?

52.A vacation rental site shows listing photos at many sizes across phones, tablets, and desktops. How should image sizes be produced?

53.A cookbook community needs search with typo tolerance, facets by cuisine and cook time, and good ranking. Postgres is the primary store. What is the usual design?

54.On a secondhand furniture marketplace, a seller edits a price and search results show the old price for about 30 seconds. What is the senior response?

55.A job board writes a posting to the database and then calls the search index. Sometimes the index call fails, and the index drifts out of date. What is the fix?

56.You need to change the text analyzer on a live product search index for an outdoor gear store. How do you do it without downtime?

57.A help desk product for many companies uses one shared search cluster for all tickets. How should tenant isolation work in search?

58.A newsletter tool must send a 2 million email campaign, and the email provider allows 500 sends per second. How should sending work?

59.An image processing queue for a print-on-demand shop keeps growing. Workers are already maxing out the database. What do you do first?

60.A school district app runs report card exports and password reset emails on one queue. Big exports delay resets by many minutes. What is the fix?

61.A language learning app must send each of 10 million users a reminder at 8 am in their own time zone. How should this be scheduled?

62.A teammate says the new queue for a parking app gives exactly-once delivery, so workers need no duplicate handling. What is the senior response?

63.Fulfilling an order at a furniture store means charging the card, reserving stock, and booking a courier, and any step can fail. How should this be built?

64.A theater sells 50,000 reserved seats at onsale. How do you prevent two buyers from getting the same seat?

65.A food festival sells 20,000 general admission tickets. Every purchase decrements one counter row, and that row becomes a bottleneck. What is the fix?

66.A prepaid transit card system updates balances in place, and support cannot explain disputed balances. What data model fixes this?

67.A physical therapy clinic system has patients, therapists, slots, and bookings, with strict constraints and transactions. Which storage fits best?

68.A water utility stores meter readings at 5,000 writes per second. Queries are by meter and time range, raw data is kept 30 days, and hourly rollups are kept for years. What design fits?

69.A journaling app must fully erase a user's personal data when they delete their account. What does that require?

70.A neighborhood forum's feed joins 5 tables on every read at 3,500 requests per second, and the database is struggling. What is the senior change?

71.A short video app precomputes timelines on write. One creator has 8 million followers, so each post triggers 8 million writes. What is the fix?

72.A trivia game's Postgres primary runs at 85 percent CPU, almost all from reads. What is the best first step?

73.A dating app reads profiles from replicas. After a user saves an edit, the profile page shows the old version. What is the fix?

74.A project management tool for businesses must shard. Tenant sizes vary hugely, and most queries stay within one tenant. What shard key fits best?

75.A wind farm telemetry store is sharded by timestamp range, and all current writes land on one shard. What is the fix?

76.A booking platform runs 40 app pods, each with a pool of 20 database connections. Postgres allows 500 connections, and pods fail to connect. What is the fix?

77.An online bookstore's orders table has 2 billion rows, but almost all queries touch the last 90 days. What helps most?

78.After sharding a sales tool by tenant, leadership wants global reports across all tenants. How should these run?

79.A building inspection app works offline for hours, and two inspectors rarely edit the same report. How should sync handle conflicts?

80.An offline-first notes app lets users create notes and link them together before syncing. How should note ids be assigned?

81.A to-do app adds items optimistically. The server may trim or reorder fields before saving. What should the client do when the response arrives?

82.Two people type in the same paragraph of a shared meeting notes document at the same time. How should edits be merged?

83.Shoppers on a sneaker store say checkout is slow, but server median latency looks fine. What should you add?

84.Which service level objective (SLO) is best for a clinic appointment booking service?

85.A tutoring marketplace has a 99.9 percent availability objective over 30 days. This month it had 30 minutes of full outage. How much error budget remains?

86.A travel booking dashboard shows average latency of 120 ms, yet users complain that searches are slow. What should you look at?

87.An on-call team for a meal kit service is paged nightly when CPU passes 80 percent, but users see no impact. What is the better alerting approach?

88.Metrics for a ride dispatch service include user_id as a label, and the metrics bill has exploded. What is the right approach?

89.After a release of a ride dispatch map, JavaScript errors spike on one browser, but stack traces are minified and unreadable. What helps most?

90.A cooking class platform serves 50 TB a month of video from origin at $0.08 per GB of egress. What is the monthly egress cost?

91.A quiz app's cloud bill is 60 percent app servers that run at peak capacity all night while traffic is low. What is the best change?

92.A legal document service stores about 15 TB of logs a month, nearly all routine success logs. How do you cut cost without losing what matters?

93.A 4-person team building a business tool needs search over tens of thousands of records per tenant. What is the sensible first choice?

94.You must move bookings from a monolith's database to a new service's database for a car rental company without downtime. What is the safest approach?

95.A bug in version 3.2 of a pet sitting app sends a malformed field, and 15 percent of users are still on 3.2. How should the API respond?

96.You need to backfill a new column across 500 million rows in a live Postgres table for a loyalty points program. How should it run?

97.An API for a returns portal adds a new status, partially_refunded. Old mobile clients crash on unknown status values. What is the plan?

98.A home insurance company is replacing its pricing engine. How do you roll out the new one safely?

99.A recipe site's recommendation service has become slow, hitting 5-second timeouts, and recipe pages hang. What is the fix?

100.A request to an insurance quote site passes through three layers, and each layer makes up to 4 attempts (1 try plus 3 retries). What is the risk when the bottom service fails?

101.A payroll software company needs a recovery point objective of 1 hour and a recovery time objective of 4 hours for a full region loss. Which design fits?

102.Traffic for a limited vinyl record release is 20 times what checkout can handle. What keeps the system up?

103.The search cluster of a craft supplies marketplace goes down. What is the best degraded behavior?

104.An HR platform has 3,000 small tenants and a few large regulated ones that require separate storage. Which isolation model fits?

105.A legal case tracker stores all tenants in shared tables with a tenant_id column. One missed WHERE clause leaked data across tenants. What is the best defense?

106.One tenant of an accounting tool runs huge exports that slow everyone else down. What is the right fix?

107.A field service app caches monthly reports under the key report:monthly, and tenants started seeing each other's reports. What is the fix?

108.European customers of a school records company require that their data be stored only in Europe. How should the system handle this?