Kotlin Engineering Blog

Ideas you will meet

Distributed Systems

Consensus, time, failure and the fundamentals of scale-out.

What we already know about this

Our own lessons on these ideas. Read them here before you start, so it makes sense.

• Low-Level Design Practice · 11 minFood deliveryRestaurants, menus, orders and couriers: an order lifecycle with three parties, and how to keep their views consistent.Read hereClose

  The problem

  Customers browse restaurants, build an order from a menu, pay and track delivery. Restaurants accept and prepare orders. Couriers pick up and deliver them.

  Entities

  100%

  Figure SDA-LLD-09-1 · Food delivery classes · © forcsguy.com

  OrderItem copies the name and price at purchase time. Later menu edits must not change old orders.

  The order lifecycle

  100%

  Figure SDA-LLD-09-2 · Order status · © forcsguy.com

  Each party may trigger only some transitions: the restaurant accepts and marks ready, the courier marks picked up and delivered, the customer may cancel early. Keep the allowed-transition table in one place.

  const next: Record<string, Record<string, string>> = {
    PLACED:    { restaurantAccepts: "ACCEPTED", restaurantRejects: "CANCELLED", customerCancels: "CANCELLED" },
    ACCEPTED:  { startPreparing: "PREPARING" },
    PREPARING: { courierPicksUp: "PICKED_UP" },
    PICKED_UP: { courierDelivers: "DELIVERED" },
  };
  function apply(order: Order, event: string) {
    const to = next[order.status]?.[event];
    if (!to) throw new Error(`${event} not allowed in ${order.status}`);
    order.status = to; order.notifyAll();      // observers: customer, restaurant, courier apps
  }
  

  Assigning a courier

  100%

  Figure SDA-LLD-09-3 · Courier assignment · © forcsguy.com

  Matching (nearest, fastest, fairest) is a strategy, as in the cab-booking lesson. Batching two orders for one courier is an extension with its own route check.

  Design notes

  • Price calculation (items, taxes, fees, discount) lives in one service used by cart, checkout and invoice.
  • Menu availability: items can sell out; check at add-to-cart and again at placement.
  • Idempotent placement: a double tap must not create two orders; use a client-generated order key.
  • Notifications use the Observer pattern on status changes.
  • Failures: payment fails (do not place), restaurant unresponsive (timeout and refund), courier cancels (reassign).

  Real-world reading

  Related reading

  • Marc Brooker's BlogMarc Brooker · brooker.co.zaBlogs
  • PagerDuty BlogPagerDuty engineers · PagerDutyBlogs
  • Chaos Monkey (source)Netflix · GitHubGitHub
  • AWS Architecture BlogAWS architects · Amazon Web ServicesBlogs

  Show 6 moreShow fewer

  • Workload isolation using shuffle shardingAmazon Builders' Library · Amazon Web ServicesBlogs
  • Scaling Live Videos to a Billion Users at Facebook - Sachin Kulkarni, Director of Engineering at FacebookFacebookYouTube
  • Delivery Hero BlogDelivery Hero engineers · Delivery HeroBlogs
  • Neural Architecture Search (NAS) for Prohibited Item Detection at MercariMercari engineers · MercariCase Studies
  • Kotlin Engineering BlogKotlin engineers · KotlinBlogs
  • Groww BlogGroww engineers · GrowwBlogs

  Open this lesson with its quiz and exercise